Introduction

Tidal wetlands cover approximately 16,000 km2 in the conterminous USA (CONUS) (Ganju et al. 2022), despite their viability in a very narrow vertical range with respect to mean sea level (Eleuterius and Eleuterius 1979). Because of their distinct response to mean sea-level changes, salt marsh monitoring is typically focused on measuring vertical changes in elevation due to organic and mineral accretion, subsidence, and compaction (Cahoon et al. 2000). Further, the “elevation capital” of a salt marsh is an indicator of marsh lifespan and resilience to sea-level rise (SLR) (Cahoon et al. 2019; Ganju et al. 2020). The Surface Elevation Table (SET) method is a standard technique used to track vertical changes in marsh elevation over annual and decadal timescales across diverse coastal vegetated habitats (Cahoon et al. 2002). Tidal wetlands that cannot maintain elevation relative to SLR, either through autochthonous or allochthonous material trap** and burial, are likely vulnerable to future increases in SLR and may submerge and/or convert to open water and unvegetated intertidal flat (Kirwan and Megonigal 2013).

Implicit in this conversion are the horizontal processes of open-water conversion and edge erosion, both of which are coupled with vertical dynamics of sea-level change and marsh elevation change (Mariotti 2020). Open-water conversion through channel and pond widening is tied to the elevation of the unvegetated area as well as inorganic sediment supply (Mariotti 2016), such that sediment-poor, low-elevation areas will tend towards runaway expansion. Therefore, the vertical and horizontal dynamics are inherently coupled and must be considered in tandem. However, it is technically infeasible to monitor elevation change across an entire salt marsh system with SETs, and methods such as airborne lidar are not accurate enough to detect annual to decadal timescale changes in elevation (Buffington et al. 2016).

Recent advances in our understanding of horizontal processes and marsh trajectory have led to consistent efforts in tracking vegetated and unvegetated cover over entire salt marsh systems. Ganju et al. (2017) identified the unvegetated-vegetated ratio (UVVR) as a useful stability metric in microtidal salt marshes that indicates horizontal vulnerability to sediment deficits and sea-level rise, with the tip** point of a neutral sediment budget and quasi-equilibrium occurring near a UVVR of 0.10–0.13 based on sediment flux measurements and coupled biogeomorphic modeling. Wasson et al. (2019) showed that runaway expansion across several systems accelerated at UVVR > 0.15. Therefore, Ganju et al. (2022) used the 0.15 value as a nominal stability threshold to identify stable and unstable wetlands, and suggest that areas exceeding this threshold may be on a deteriorating trajectory. Ganju et al. (2022) also developed a CONUS database of the UVVR for the 2014–2018 period using Landsat imagery (Vermote et al. 2016), which quantifies within-pixel fractional vegetated cover and UVVR at 30-m horizontal resolution. This dataset provides an opportunity to assess the spatial variability in horizontal vulnerability and the relationship with vertical trajectory as measured by SETs.

Prior work combining the UVVR with elevation demonstrated a tendency for low-elevation marshes to have higher UVVR (indicating vertical and horizontal vulnerability), while higher-elevation marshes tended towards lower UVVR (Ganju et al. 2020). This relationship was consistent across four large estuarine systems; however, that analysis did not consider elevation change or relative tidal elevation. The former is more relevant in light of sea-level rise, while the latter enables comparison across diverse systems. In this study, we incorporate the SET trend analysis from Ladin and Moorman (2022) across several southeastern US coastal National Wildlife Refuges (NWR) to test two hypotheses. Firstly, we hypothesize that the SET location is representative of the overall system in terms of horizontal vulnerability. Secondly, we hypothesize that if the tidal wetland exhibits a positive vertical trajectory and high relative tidal elevation, then the marsh system should also demonstrate horizontal stability and integrity.

Methods

Site Description and Net Elevation Change

All sites are located within minimally altered coastal wetlands on NWRs (hereafter “refuges”) of the southeastern USA and are described in detail by Ladin and Moorman (2022). We considered the 18 sites that were sampled with consistent methods: these include freshwater emergent wetlands, estuarine wetlands, and emergent wooded (pocosin) wetlands across North Carolina, South Carolina, Georgia, and Florida (Fig. 1). Due to the distinct hydrologic and vegetative nature of pocosin and forested wetlands, we removed three sites for our analysis. The methods for elevation change and sea-level rise estimates are also detailed by Ladin and Moorman (2022). Standard error and confidence bounds in elevation change and sea-level rise, respectively, are taken from Ladin and Moorman (2022; Table S1).

Fig. 1
figure 1

SET and refuge locations, with general wetland type indicated; Alligator River site denoted by All. Riv

Full size image

Representativeness of SET Site

Comparison of the fractional vegetation cover and UVVR with elevation change requires the selection of the pixels within the UVVR dataset that correspond with the SET location. We evaluated UVVR at four spatial scales to assess how representative the SET site was. At the finest areal level, we aggregated the 3 × 3-pixel footprint (9 pixels at approximately 30 × 30 m, or 8100 m2) surrounding the SET location as the “site” UVVR (Fig. 2). We chose to incorporate more than the single 30 × 30 m pixel as the SET may be located on the edge of a pixel, and we assumed that SET site selection initially considered the setting across an area broader than the point itself. The second areal footprint is the National Wetland Inventory (NWI) polygon that the SET is within (Fig. 2). NWI polygons are defined based on visual interpretation and observations of habitat types; therefore, their boundaries are approximate and the areal footprint can be variable. For this study, we are only considering emergent wetland classes; therefore, the individual polygon considered will always be classified as emergent wetland habitat.

Fig. 2
figure 2

Example of SET site, and UVVR footprints for the polygon, emergent wetlands, and tidal wetland areas, for Mackay Island National Wildlife Refuge, North Carolina. Site footprint consists of a 3 × 3-pixel area with SET site at center (not shown). Each successively larger footprint includes the preceding, smaller footprint (i.e., tidal wetland footprint includes polygon and emergent footprints in addition to blue areas indicated). For example, the refuge layer includes the purple, blue, green, and red layers; the tidal wetland layer includes the blue, green, and red layers; and the emergent layer includes the green and red layers. Basemap courtesy of ESRI

Full size image

The third areal footprint is the entire emergent wetland class for the refuge (which includes three subclasses), and includes all emergent wetland polygons within the refuge boundary (Fig. 2). Lastly, we consider the tidal wetland areal footprint across the entire refuge, which combines all estuarine intertidal subsystems, and all subsystems in the classification that are marked with a tidal modifier (NWI exclusive mask; Ganju et al. 2022). This footprint will yield the broadest possible characterization of UVVR for the refuge, but typically will not increase the overall area by a large amount (Fig. 2).

Across each footprint, the CONUS-wide fractional vegetation and UVVR product of Couvillion et al. (2021) was used. The 2014–2018 composite product is used to remove annual variability and give the average condition over that time period; the vegetated fraction across all pixels within the footprint was multiplied by the footprint area to calculate a total vegetated area (Av), which was then subtracted from the total area (At) to calculate the unvegetated area (Auv), and the UVVR was calculated as Auv/Av. We then compare each site’s elevation change and net elevation change relative to SLR (i.e., elevation change minus SLR) with the UVVR across these four footprints. Maximum root-mean-square error in the vegetated fraction was estimated as 19% and propagated at all sites (Ganju et al. 2022).

We also utilize the CONUS-wide relative tidal elevation (Z*, Holmquist and Windham-Myers 2022) to investigate relationships with elevation change and UVVR. This metric, defined as the elevation above mean sea level (MSL) normalized by the mean high-water level (relative to MSL), provides a complementary spatially integrative assessment of the elevation capital across each refuge; prior work across northeastern U.S. marsh complexes indicated strong coherence between elevation and UVVR (Ganju et al. 2020). The Z* data used here span the CONUS at 30-m resolution and cover estuarine marshes as indicated by the Coastal Change Analysis Program regional land cover (NOAA 2022). These data were extracted at both the SET site as well as aggregated over the tidal wetland footprint for comparison with elevation change and the UVVR. Uncertainty in Z* is covered in detail by Holmquist and Windham-Myers (2022) and arises from both errors and/or biases in lidar measurement (for elevation) and uncertainty in tidal datums and extrapolation used to estimate mean sea level and mean high water.

Results

Representativeness of SET Site

We compared the “site UVVR” (3 × 3 pixel) surrounding the SET location and the UVVR calculated across the footprints of the associated NWI polygon, entire emergent vegetation layer, and entire tidal wetland layer across the corresponding refuge (Tables 1 and 2; Fig. 3). In general, the site UVVR tracked with the expanded footprint UVVRs well, with the correlation maximized for the emergent wetlands layer footprint (r2 = 0.86). At two sites, the site UVVR was over the nominal UVVR stability threshold of 0.15, and one or more of the expanded footprint UVVRs were under the threshold. At Harris Neck NWR, the NWI polygon which contains the SET site is a narrow swath of intact emergent wetland that is bordered by degraded wetlands and mudflat on the seaward edge, and upland on the landward edge, yielding a relatively lower UVVR. Conversely, the overall emergent wetland and tidal wetland layers show substantially higher UVVR and are better correlated with the local site UVVR. At Mackay Island NWR, the site UVVR is elevated due to the SET’s proximity to open water, such that pixels adjacent to the site include water in the calculation of the UVVR. All three expanded footprint UVVRs at that site are substantially lower and under the stability threshold. Overall, the SET site selections are representative of the salt marsh condition across the refuges, with the aforementioned exceptions. Given the qualitative nature of NWI polygon selection, and the high correlation between the emergent and tidal wetland footprint UVVRs, the subsequent analyses compare the emergent wetland UVVR, which integrates a larger wetland area than the site or polygon UVVR, with elevation change.

Table 1 Refuge name, area of each UVVR footprint, and UVVR values. Site footprint areas are nominally the area of nine 30 × 30 m pixels or 0.0081 km2. National Wetlands Inventory (NWI) data were used to delineate footprints (https://www.fws.gov/program/national-wetlands-inventory)
Full size table
Table 2 Coefficient of determination (r2) between UVVR values across footprints
Full size table
Fig. 3
figure 3

Comparison of site UVVR values (3 × 3-pixel area) with expanded UVVR footprints at each refuge; polygon refers to the NWI polygon that contains the SET site, emergent refers to the emergent wetland layer, and tidal wetlands refers to the tidal wetland layer (see Fig. 2). Inset shows comparison with threshold UVVR value of 0.15 indicated by dashed lines

Full size image

Elevation Change vs. UVVR

Elevation change was positive at all sites (Fig. 4a; Table 3). A power fit to the relationship yields a moderate correlation (r2 = 0.51); however, there is no a priori expectation that elevation change and vegetative cover should be universally correlated. Nonetheless, the highest rates of elevation change occur at sites with the highest vegetative cover and lowest UVVR, whereas low rates of elevation change are seen across a range of vegetative cover and UVVR. Above the UVVR threshold of 0.15, the maximum rate of elevation change is approximately 2 mm/year, and under that threshold, the maximum rate is 7 mm/year, with the majority of sites above 2 mm/year.

Fig. 4
figure 4

Comparison of UVVR over tidal wetland footprint and (A) elevation change and (B) net elevation change (elevation change – local sea-level rise) across 15 refuges. UVVR threshold of 0.15 indicated by vertical dashed line. Elevation and sea-level rise error bars taken from Ladin and Moorman (2022) and reported in Table S1

Full size image
Table 3 Refuge name, UVVR, and Z* pixel-averages over the tidal wetland footprint, and SET site data for marsh elevation change. Sea-level rise is taken from Ladin and Moorman (2022; Table S1), and net vertical change is computed as the net difference between the elevation change and sea-level rise. UVVR values differ from Table 1 due to inclusion of within-polygon null pixels reported in Table 1
Full size table

Net Elevation Change vs. UVVR

Net elevation change was positive at six sites, three of which were oligohaline, and three of which were salt marsh (Fig. 4b; Table 3). The salt marsh sites were all in North Carolina, within the wind-driven Albemarle-Pamlico Sound system that is weakly influenced by tides (Wells and Kim 1989). All six sites with net positive change have UVVRs below the 0.15 threshold, ranging from a low of nearly zero at Waccamaw NWR and a high of 0.08 at Swanquarter NWR, indicating that vertical gain without horizontal integrity was not observed across these refuges. The remaining nine sites had negative net elevation change.

Four of the sites with negative change fell below the 0.15 threshold, and the remaining five were above. In those cases, the UVVR ranged from a low of 0.08 at Mackay Island NWR to a high of 2.8 at Pinckney Island NWR. Every salt marsh site within South Carolina and Georgia had negative net elevation change and UVVR exceeding 0.15.

Elevation Change, UVVR, and Z*

Elevation change (both total and net relative to SLR) did not show noticeable trends with Z* at either the site or tidal wetland footprint level, even when partitioned by wetland type (not shown). However, Z* and UVVR (Table 3) had strong correlation over the tidal wetland footprint when partitioned between salt marshes (r2 = 0.64; power fit of Z* = 0.58 UVVR−0.48) and oligohaline (r2 = 0.80; linear fit of Z* =  − 4.15 UVVR + 0.8) wetlands (Fig. 5). Based on these best fits, the UVVR threshold of 0.15 is crossed at Z* = 0.18 for oligohaline marshes, and 1.4 for salt marshes, considering the data points straddling a UVVR of 0.15 (and not a fitted curve) suggests a threshold Z* of 1 for salt marshes (no oligohaline sites had UVVR > 0.15). Note that both Mackay Island and Currituck NWRs are in weakly tidal areas (Ablemarle-Pamlico Sound) that are subject to large wind-induced water level fluctuations (Luettich et al. 2002); therefore, Z* estimates should be interpreted with caution.

Fig. 5
figure 5

Relationship between UVVR and relative tidal elevation (Z*) over the tidal wetland footprint at each refuge. Power and linear fits to the salt marsh, and oligohaline sites respectively yields Z* = 0.58 UVVR−0.48 (for salt marshes, r2 = 0.64) and Z* =  − 4.15 UVVR + 0.80 (for oligohaline marshes, r.2 = 0.80). Bounds represent 25th and 75th percentiles of the distributions of pixels within each tidal wetland footprint; 75th percentiles of UVVR for sites Wassaw and Pinckney are 4.3 and 10.9, respectively; 75th percentile of Z* for site Cedar Island is 4.2. Vertical dashed line is the UVVR threshold of 0.15

Full size image

Discussion

Coherence Between Vertical Trajectory and Horizontal Integrity

Across all wetland types, this analysis shows that intact vegetative cover is a prerequisite for vertical stability. None of the sites with UVVR above 0.15 had positive net vertical change, indicating that a stable marsh platform and positive net vertical accretion are either causally linked or correlated. A causal link would most likely arise through hydrologic/biogeochemical mechanisms that limit vertical growth when vegetative cover above the substrate falls below a certain threshold; this may operate in tandem with a correlative link whereby sea-level rise and position in the tidal frame both limit vertical accretion and contribute to horizontal deterioration through other mechanisms, such as sediment export and pond/channel widening (Luk et al. 2023).

With regard to salt marshes, no sites with UVVR > 0.15 exceeded 2 mm/year of elevation change; this rate is well below future scenarios of sea-level rise for the southeastern USA (Sweet et al. 2017). The two sites under the UVVR threshold, and with negative net elevation change, are St. Marks and Pea Island NWRs. Both of the SET sites in these refuges are situated at the landward edge of the salt marsh near the upland boundary, in microtidal systems. It may be that these sites are minimally influenced by sea-level rise processes due to their location and are not yet responding to a modified position in the optimal biomass growth-inundation relationship (Kirwan and Guntenspergen 2010).

No salt marsh sites in South Carolina (SC) or Georgia (GA) have net elevation gain greater than zero, and all are below a Z* threshold of 1 m and above a UVVR threshold of 0.15. Ganju et al. (2022) noted that marsh plains in both SC and GA tended to reflect high UVVR values likely due to sparsely vegetated plains. The high UVVR in SC and GA salt marshes, which is coherent with negative rates of net elevation change suggests a connection between sparse vegetation and an inability to build vertically, whether due to limited sediment trap** or preservation of belowground production and growth. The Landsat-based UVVR method assigns a continuous value to vegetative cover (Couvillion et al. 2021), as opposed to a “binary” vegetated/unvegetated value; therefore, a sparsely vegetated plain within a 30 × 30 m pixel could have the same UVVR value as a densely vegetated plain with ponding. Salt marshes in the SC and GA refuges studied here have sparsely vegetated plains, coincident with low relative tidal elevation, indicating that conversion to open water or bare sediment may be actively occurring. Oligohaline sites appear more horizontally intact with higher elevation change overall, and the only sites in SC and GA with net elevation gain greater than zero and below the UVVR threshold are oligohaline.

The comparison between relative tidal elevation and the UVVR reinforces the importance of elevation capital on overall wetland stability. Ganju et al. (2020) speculated that internal and external forces of deterioration (flooding, edge erosion, tidal export of sediment) are more pronounced at lower elevations, leading to a tendency towards open-water conversion. However, the strong relationship with Z* perhaps further indicates a more intrinsic coupling between vertical and horizontal stability. For example, does increased UVVR due to an initial perturbation alter hydrology within the substrate and overall system that accelerates elevation loss, and induces more open-water conversion? Oligohaline marshes and salt marshes do appear to have differing elevation-UVVR thresholds, with oligohaline marshes maintaining vegetative cover despite being at lower elevations. Additionally, taken as a group, oligohaline marshes appear to be kee** pace with SLR slightly better than salt marshes (at least at discrete SET sites) in terms of net elevation change. This may be due to vegetation differences, distance from sediment sources, accretion rates, or hydrological differences influencing conditions within the marsh substrate. However, Krauss et al. (2018) suggest oligohaline and tidal freshwater wetlands may experience rapid conversion to salt marsh in the future.

Incorporating Spatial Metrics into Marsh Restoration Assessments

SETs are an intensive measurement technique that provides the marsh vertical response to sea-level rise with high fidelity at specific points within the marsh landscape. Given the difficulty of deploying SETs across large expanses at sufficient spatial resolution and the delay between installation and generation of robust data, spatially integrative metrics can be used in tandem. One use is to provide a rapid assessment of where SET measurements may be essential or trivial. For example, the UVVR—elevation change relationship (Fig. 4b) indicates that elevation change does not keep pace with sea-level rise at sites where the UVVR exceeds the threshold value of 0.15. Therefore, prioritization for restoration can readily focus on these areas with techniques such as sediment augmentation, runneling, or revegetation designed to increase vegetated area, decrease the UVVR, and perhaps contribute to positive vertical growth of the plain. Areas with UVVR less than 0.15 would likely not benefit from intervention given the lack of techniques available to solely increase the elevation of already vegetated plains, unless there is an external stressor such as a tidal restriction or sediment deficit due to anthropogenic actions that can be mitigated.

Furthermore, the combination of Z* and the UVVR represents a synthesis of two spatially aggregated metrics that illustrate the coupled vertical and horizontal status across entire complexes. If stability thresholds in both Z* and the UVVR can be established across or within individual complexes, map** the distribution of pixels that fall into a simple four-box decision matrix (Fig. 6; i.e., high Z*/low UVVR, low Z*/high UVVR, low Z*/low UVVR, high Z*/high UVVR) can help identify vulnerable areas as well as restoration techniques. For example, sites with high Z* and low UVVR may be candidates for land acquisition, while sites with high Z* and high UVVR may benefit from runneling and revegetation given the existing elevation capital, perhaps moving the site away from a trajectory of decline.

Fig. 6
figure 6

Conceptual model of vertical and horizontal coherence. Sites in the top left quadrant exhibit vertical and horizontal stability simultaneously (Waccamaw NWR shown; Currituck, Savannah, Alligator River, Cedar Island, and Swanquarter also fall in this quadrant); sites in the bottom left quadrant do not show horizontal deterioration despite not maintaining elevation relative to sea-level rise (Pea Island NWR shown; ACE Basin, Mackay Island, and St. Marks also fall in this quadrant), and may either be rapidly submerging or transitioning to the bottom right quadrant of coherent vertical and horizontal deterioration (Pinckney Island NWR shown, includes all SC and GA salt marsh sites). The top right quadrant, not observed in this study, represents marsh plains that are horizontally degrading but maintaining positive vertical change in response to sea-level rise. This could arise in a high wave energy environment with fragmented marsh. Relative tidal elevation threshold of Z*crit represents a possible tip** point in the Z*-UVVR relationship as suggested by Fig. 5. Basemap imagery from ESRI

Full size image

Coupled Vertical and Horizontal Response to Sea-Level Rise

Comparing elevation change and horizontal integrity provides insight into the response of coastal wetlands to external stressors. Marshes with coherent vertical and horizontal trajectories illustrate the classical transgressive response to sea-level rise. Specifically, marshes with net vertical gain and horizontal stability may represent growing features in a transgressive landscape that is translating landward under sea-level rise (i.e., “classical positive trajectory”) or seaward-expanding features in a regressive landscape (Miller et al. 2022). Conversely, marshes with net vertical loss and horizontal instability follow a conceptual model of horizontal disintegration as the marsh plain submerges, also a transgressive response to sea-level rise (“classical negative trajectory”). In a scenario with net vertical loss with no evidence of horizontal deterioration, this may indicate resilient vegetative communities that maintain cover but are not accreting rapidly enough to offset subsidence (“submergence trajectory”). Lastly, there is potential for a “cannibalistic trajectory,” whereby a marsh gains vertically while deteriorating horizontally, suggesting scavenging of the eroding plain to aid in accretion (Luk et al. 2021). None of the sites in this study fell into this category; however, it may be possible to observe this behavior in exposed environments where high wave energy and large tidal range may erode fringing marsh and deposit sediment during high tides, with corresponding marsh plain break-up. Alternatively, highly fragmented areas with significant sediment mobilization, such as the Mississippi Delta region, may demonstrate high rates of vertical accretion (though subsidence may be relatively high), along with horizontal deterioration. SETs are not typically installed in rapidly degrading areas; however, sediment accretion measurements near erosive shorelines may support this observation (Hopkinson et al. 2018).

The relationships between net elevation change and UVVR, as well as Z* and UVVR, appear to vary based on the environment, specifically oligohaline versus salt marshes. These relationships may illustrate the path by which vertical and horizontal dynamics change with sea-level rise as oligohaline marshes convert to salt marsh. For example, oligohaline marshes in this study had generally higher elevation change (both absolute and net) along with lower UVVR as compared to salt marshes. As salinity and water level increase with sea-level rise, oligohaline marshes may decrease their pace of vertical accretion (either due to salinity stress or position in the tidal frame) and simultaneously convert to open water in places, thereby shifting both down and to the right in the conceptual model (Fig. 6). There may be a commensurate shift in the relationship between Z* and UVVR, with oligohaline marshes maintaining a similar relative tidal elevation with sea-level rise but deteriorating horizontally. This is speculative, however, given variability in future tidal range across systems. Determining future tidal range with sea-level rise and marsh loss requires coupled geomorphic-hydrodynamic models of entire estuarine systems (Donatelli et al. 2018).

Conclusion

Coastal wetlands adjust their vertical position in response to tidal and sea-level changes, thereby requiring intensive monitoring of elevation change to assess their response to external forcings. SETs are an important component of this monitoring, and complementary spatially aggregated metrics can be used to evaluate the representativeness of SET locations as well as the coherence between vertical and horizontal processes. Established thresholds for the UVVR, a metric of horizontal stability, do not uniformly predict whether or not a location is kee** pace with sea-level rise; however, all sites in this study kee** pace do fall below the critical threshold of 0.15. This suggests that intact vegetative plains are a prerequisite for vertical stability in light of sea-level rise. Spatially aggregated metrics such as the UVVR and relative tidal elevation maps are important tools for rapid assessment and identifying areas that may require intensive monitoring through SETs and other techniques that can decipher instability mechanisms at the site-specific scale.