Skip to main content
NCBI home page
EPA Author Manuscripts logoLink to EPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Jan 1.
Published in final edited form as: Remote Sens Appl. 2023 Jan 1;29:1–11. doi: 10.1016/j.rsase.2022.100910

Improved mapping of coastal salt marsh habitat change at Barnegat Bay (NJ, USA) using object-based image analysis of high-resolution aerial imagery

Johannes R Krause a,b,*, Autumn J Oczkowski c, Elizabeth Burke Watson b
PMCID: PMC10208303  NIHMSID: NIHMS1895626  PMID: 37235064

Abstract

Tidal wetlands are valued for the ecosystem services they provide yet are vulnerable to loss due to anthropogenic disturbances such as land conversion, hydrologic modifications, and the impacts of climate change, especially accelerating rates of sea level rise. To effectively manage tidal wetlands in face of multiple stressors, accurate studies of wetland extent and trends based on high-resolution imagery are needed. We provide salt marsh delineations for Barnegat Bay, New Jersey, by means of object-based image analysis of high-resolution aerial imagery and digital elevation models. We performed trends analyses of salt marsh extent from 1995 to 2015 and estimated drivers of marsh area change. We found that in 1995, 8830 ± 390 ha were covered with marsh vegetation, while in 2015 only 8180 ± 380 ha of salt marsh habitat remained. The resulting net loss rate of 0.37% yr−1 is equivalent to historic loss rates since the 1970s, indicating that despite regionally accelerating relative sea level rise and purported eutrophication, salt marsh loss rates at Barnegat Bay remain steady. The main drivers of salt marsh loss are excavations for mosquito control (409 ha), edge erosion (303 ha) and ponding (240 ha). Upland migration of salt marsh did not completely mitigate these losses but accounted for a gain of 147 ha of tidal marsh habitat. The methodology presented herein yielded accurate salt marsh delineations (>90%) and trend detection (85%), outperforming low-resolution wetland delineations used in coastal management. This study demonstrates the suitability of high-resolution imagery for the detection of open water features. For the purposes of salt marsh change detection and the identification of change drivers, management and conservation agencies should make use of high-resolution imagery whenever feasible.

Keywords: Salt marsh, Trend detection, High-resolution, Object-based, Ponding, Mosquito control, Eutrophication

1. Introduction

Salt marshes are among the most productive coastal ecosystems and valued for the ecosystem services they provide (Barbier et al., 2011), such as coastline protection via wave attenuation and sediment stabilization (Koch et al., 2009), enhancement of commercial and recreational fisheries through the provision of food and habitat (Deegan et al., 2002), and their role in alleviating coastal eutrophication as sites of increased nutrient sequestration and removal (Velinsky et al., 2017). In addition, tidal marshes store significant volumes of carbon in their soils, and thus play an important role in climate change mitigation (Duarte et al., 2004). To support management of these valuable natural resources, assessments of wetland status and trends are carried out on the national (Dahl and Stedman 2011) and global levels (Davidson et al., 2018).

Despite long-standing efforts to inventory salt marshes and other wetlands, past research produced conflicting results. For example, a recent review found that estimates of the global extent of wetland area were steadily increasing from 1979 to 2015 (Davidson et al., 2018). In contrast, studies focused on wetlands trends analysis routinely report losses (Lotze et al., 2006; Macreadie et al., 2021). Contradicting reports also occur on a regional scale. In the U.S., the National Wetlands Inventory (NWI) reported a decline of loss rates since the 1950s and a net gain of wetland area at the beginning of the 21st century, including in the U.S. mid-Atlantic and Northeast (Dahl and Stedman 2011). However, individual trend analyses for tidal wetlands in this region report continuing coastal marsh fragmentation and loss. For example, analyses conducted for a ‘State of the Estuary’ report indicate that Delaware Bay has lost an acre per day of tidal wetlands between 1996 and 2006 (PDE 2012), and a recent analysis of Long Island tidal wetlands shows a loss rate of ca 15% between the early 1970s and early 2000s (Cameron Engineering and Associates, 2015). Similar rates and patterns of losses have also been estimated for Chesapeake Bay and southern New England (Kearney et al., 2002; Smith 2009; Watson et al. 2014, 2017), suggesting that these patterns and trends are regionally widespread.

One explanation for the discrepancies in reported wetland trends is the improvement of mapping technologies available to investigators, particularly the rise of high-resolution (<5 m) satellite imagery (Davidson et al., 2018). Recent studies have taken advantage of an increased availability of high-resolution imagery to show that wetland mapping produces higher accuracy results with increasing imagery resolution, due to reduced spectral mixing and a better accordance of pixel size and the scale of features of interest (Aplin 2006; McCarthy et al., 2015; Turpie et al., 2015). For global inventories, this meant an improved detection of wetlands, consequently leading to larger estimated wetland extent in more recent studies, albeit no actual increase of wetland extent occurred (Davidson et al., 2018). This phenomenon may also explain site-specific inconsistencies in wetlands trend analyses. For instance, Barnegat Bay in New Jersey lost 27% of salt marshes between 1888 and 1972 (Lathrop and Bognar 2001). Since then, an additional 12% of salt marsh habitat was lost, according to an estimate based on 2012 NJ Land Use Land Cover data (BBP 2016). Yet, using high-resolution (1 m) orthoimages from 2015 for change detection in three Barnegat Bay salt marsh sites, Watson et al. (2017) derived a loss of only 9.7% since 1975. This discrepancy in salt marsh area and change rate estimates is typical for analyses carried out at different spatial resolutions.

In addition to emergent wetland vegetation, improved imagery resolution can also enhance the detection of small-scale non-wetland features, such as ponds, bare spots, channels, and ditches. For example, Thomsen et al. (2022) tracked wetland colonization by individual plant patches at a recently restored California salt marsh, Powell et al. (2020) identified individual ponds on the marsh platform of Barnegat Bay, and Timm and McGarigal (2012) mapped vegetation and sand patches using 1-m orthophotography of Cape Cod National Seashore salt marshes. These studies used remote sensing to investigate features and processes operating on spatial scales smaller than the pixel size of satellite sensors commonly used for vegetation monitoring (e.g., Landsat, 30 m px−1), such as vegetation patches, disturbance, or marsh habitat use. More generally, salt marsh platforms can be heterogenous and fragmented, and imagery with higher resolution was found to map salt marsh extent more accurately on the landscape scale (McCarthy et al., 2015).

In the U.S. mid-Atlantic and Northeast regions, conversions of tidal wetland habitat to open water are often associated with pond formation and expansion, channel widening, vegetation dieback, and edge erosion, all of which are symptoms of ‘marsh drowning’ (e.g., Blum and Roberts 2009). Potentially operating on small spatial scales, these processes of tidal wetland loss and fragmentation are thought to occur where salt marsh accretion rates fail to keep up with relative sea level rise (SLR) and may be exacerbated by nutrient pollution (Kennish 2001a; Adam 2002; Krause et al., 2020). High nutrient levels may cause a shift from below to above-ground production and contribute to a loss of soil volume (Darby and Turner 2008), structural changes in soil properties (Wigand et al., 2014) and decreased soil cohesion (Turner et al., 2009). Increased nutrient levels can ramp up decomposition, leading to high levels of phytotoxic sulfides (Watson et al., 2014). Finally, high nutrient levels may contribute to large areas of macroalgal wrack on the marsh platform, which can cause temporary to permanent die-back areas (Kirwan et al., 2008; Wasson et al., 2017).

Excess nutrients can play a subtle yet important role in dieback, exacerbating the effects of climate change in vulnerable coastal marshes. In addition to preventing and reducing eutrophication, coastal management can support salt marsh stability in face of SLR by allowing the natural process of upland migration to unfold. Kirwan et al. (2016) suggest that with SLR, marshes have the potential to expand in area, because migration into the upland is more sensitive to rising sea levels than is edge erosion. This notion is supported for salt marshes at Chesapeake Bay, where the area of upland conversion to salt marsh was larger than that of marsh loss through drowning since the 19th century (Schieder et al., 2018). However, this process of marsh extent stabilization can be hindered by anthropogenic coastline modification (Kirwan et al., 2016), where migration barriers result from coastal development (Schieder et al., 2018), or where wetland hydrology is impacted by diking (Wasson et al., 2013).

Since signs of marsh drowning and wetland loss are widespread in the Eastern U.S., there is a need for tidal wetlands trends analyses based on high-resolution imagery able to resolve the associated processes that may operate on small spatial scales. Still, the use of high-resolution imagery also brings challenges, namely the exacerbation of georectification errors due to a reduced margin of error for smaller pixel sizes (Timm and McGarigal 2012) and oversampling resulting from small-scale heterogeneity of features belonging to the same class (Aplin 2006). Previous studies approached the latter problem by aggregating pixels into image objects via object-based image analysis workflows (Pierce 2015; Powell et al., 2020).

This study addresses the need for trends analyses of salt marsh habitat that adequately capture the complex marsh topography, including channels, ponds, and ditches, for Barnegat Bay, New Jersey. Located in a region where rapid SLR (Haaf et al., 2021) and eutrophication co-occur (Kennish and Fertig 2012), accurate high-resolution geospatial analysis is needed to inform coastal management efforts and to guide effective SLR planning (Weis et al., 2021). To determine rates of marsh areal change in Barnegat Bay from 1995 to 2015, we conducted a trend analysis using land cover classifications of aerial imagery and digital elevation models produced via object-based image analysis. To further assess drivers of salt marsh change and identify potential signs of marsh drowning, we performed probability sampling of 500 locations where marsh change occurred and assigned the change classes as edge erosion, ponding, mosquito control excavation, channel widening, upland migration, or revegetation of previously bare or ponded marsh surface. We contrast our results with previous salt marsh change estimates based on coarser imagery (30 m px−1) and discuss the prevalence of salt marsh change drivers in the context of SLR and nutrient management.

2. Materials and methods

2.1. Study system

Barnegat Bay is a shallow lagoon-type estuary of the mid-Atlantic, with a surface area of approximately 280 km2 (Kennish 2001b) of which 92 km2 comprised salt marsh habitat in 2012 (BBP 2016). A barrier island complex separates the bay from the Atlantic Ocean, with two inlets allowing for semidiurnal tidal exchange. Tidal ranges are highest at the inlets (1.4–1.5 m), but greatly reduced within the bay (0.2–0.3 m) and the flushing time for the estuary was estimated between 27 days in January and 71 days in July (Kennish 2001b, and references therein). While early reductions in Barnegat Bay salt marsh area were mainly due to coastal development and fill, this activity largely ceased following the Wetlands Act of 1970 (Lathrop and Bognar 2001). Other factors drove salt marsh loss at Barnegat Bay over the following 50 years, among them were losses due to edge erosion (Leonardi et al., 2016), excavations for mosquito control (Powell et al., 2020), and accretion deficits in face of rising sea levels (Haaf et al., 2021).

2.2. Data acquisition

Data for the classification of coastal land-cover of Barnegat Bay (Fig. 1) were downloaded from the USGS Earth Resources Observation and Science (EROS) Center via the USGS Earth Explorer website (Table 1). The spatial data comprised high-resolution orthoimages acquired in April 1995 (color-infrared, 3 bands, 1 m resolution) and June 2015 (4 bands, red, green, blue, near-infrared, 0.3 m resolution), as well as digital elevation models (DEMs) from 2000 (Elevation Derivatives for National Applications, 10 m resolution,https://doi.org/10.5066/F7TD9VTQ) and 2015 (Coastal National Elevation Database, 1 m resolution, https://doi.org/10.5066/F7Z60MHJ).

Fig. 1.

Fig. 1.

Open in a new tab

The study region of Barnegat Bay, New Jersey. Approximate salt marsh area is indicated in green. The inner map shows the position of Barnegat Bay (pink frame) on the U.S. east coast.

Table 1.

Geospatial data sources for the classification of Barnegat Bay salt marsh habitat, courtesy of the U.S. Geological Survey. Tidal stage relative to mean lower low water estimated from time of image acquisition using tide predictions from NOAA Tides & Currents (tidesandcurrents.noaa.gov) for gauges at Tuckerton, Waretown, and Tom’s River.

Type 1995 Imagery 1995 Elevation 2015 Imagery 2015 Elevation
Aerial photo – digital orthophoto quadrangle Bathymetric model Aerial photo – orthoimgaery Topobathymetric model
Acquisition date 04/1995 2000 06/2015 2014
Tidal stage −0.1 – 0.1 m 0.0–0.1 m
Bands CIR (3) Mono (1) RGB + NIR (4) Mono (1)
Resolution 1 m 10 m 0.3 m 1 m
Open in a new tab

2.3. Image processing and analysis

Digital elevation models were clipped to the area of interest and resampled to 1-m cell size in ArcMap 10.2.2 (Esri, West Redlands, CA, USA) (Fig. 2). Aerial imagery orthomosaics were clipped to the same extent and stacked with the DEMs to generate a five-band raster file for 2015 (red, green, blue, near-infrared, elevation; 1-m cell size) and a four-band raster file for 1995 (red and near-infrared combined in band 1, 1-m cell size). These raster files were tiled into four smaller subsets (tiles 1–4) to aid processing and subsequently imported into eCognition Developer 9 (Trimble, Sunnyvale, CA, USA) for object-based image analysis. After multi-resolution segmentation (shape = 0.1; compactness = 0.2), a supervised classification process was used to assign each image object to the classes ‘Open Water,’ ‘Upland,’ ‘Salt Marsh,’ or ‘Unvegetated’. Class assignment was based on object mean values for bands 1–5, maximum difference, hue, or brightness via two separate rulesets (one ruleset for each time point, see Figs. S1 and S2). Obvious classification errors that were identified upon visual inspection of the initial classification (e.g., salt marsh class assigned to open water object) were manually corrected before classification results were exported as raster files (Olofsson et al., 2014). A change map was generated from the 1995 and 2015 classifications, assigning cells to the four classes of marsh loss, marsh gain, stable non-marsh, and stable marsh.

Fig. 2.

Fig. 2.

Open in a new tab

Schema of geospatial imagery processing and classification workflow. Green shapes indicate imagery files, blue shapes indicate classifications.

2.4. Areal trends and accuracy assessment

To estimate the total area of change classes with confidence intervals, a reference dataset of 1600 points was allocated to the four change classes in a stratified random design. The number of required reference points was estimated setting a target standard error of overall accuracy of 0.01 and expected class-specific user’s accuracy of 0.6 for change classes and 0.8 for stable classes (Olofsson et al., 2014). To assign the 1600 random points to change classes, we used an intermediate of proportional and equal allocation to balance the needs for simultaneous estimations of area (ideally proportional) and user’s accuracy (ideally equal). This resulted in 600 reference points allocated to the class stable marsh, 500 points to stable non-marsh, 200 points to marsh gain and 300 points to marsh loss.

Labeling of the reference dataset via photointerpretation was based on the higher resolution imagery for 2015 (30 cm) and 1-m orthophotos for 1995 in addition to inspection of Google Earth imagery from 1995, 2002, 2013, and 2016. A confusion matrix was constructed by extracting the change map classes at reference point locations and expressing them as proportion of total map area. Areal trends with 95 %-confidence intervals were then calculated from the error matrix following Olofsson et al. (2014). The total marsh extent in 1995 was calculated by summing the estimated area of stable marsh and the area of marsh loss, while the total marsh extent for 2015 is the product of stable marsh area and the area of marsh gain. Confidence intervals for the 1995 and 2015 marsh extent estimates were derived by adding standard errors of the corresponding change classes. In addition, confusion matrices were created by intersecting the reference points with our 1995 and 2015 classifications to estimate classification accuracies of each map (Table S1; S2).

To compare the accuracy of our classification product to wetland delineations commonly used by coastal management practitioners, we downloaded NOAA Coastal Change Analysis Program (C-CAP) data for our study region (https://coast.noaa.gov/digitalcoast). C-CAP data were available for 1996 and 2016, corresponding well to the timepoints of our classification and reference products. The area of estuarine emergent wetlands was extracted for comparison to other Barnegat Bay wetland extent estimations. We performed accuracy assessments using the 1600 reference points described above and computed overall accuracy using confusion matrices for C-CAP wetland delineations of 1996, 2016, and 1996–2016 change trajectories (Table S3; S4; S5).

2.5. Drivers of change

Drivers of changes in Barnegat Bay salt marsh extent were estimated by randomly placing 500 points into the change map polygons classified as salt marsh gain or salt marsh loss. These were labeled following the same procedure as described for area and accuracy estimation above, using the classes ‘edge erosion’, ‘ponding’, ‘mosquito control’, and ‘channel widening’ for conversions of marsh to non-marsh, as well as ‘upland migration’, ‘pond revegetation’, and ‘bare revegetation’ for conversions of non-marsh to marsh. We then calculated the relative proportions and change areas of these drivers of marsh gain and loss.

3. Results

3.1. Mapping results and accuracy

Salt marsh habitat was mapped for the entirety of the Barnegat Bay estuary using high-resolution imagery from 1995 to 2015 and an object-based image analysis workflow. This mapping effort resulted in areal estimates for stable marsh (6968 ± 258 ha), marsh loss (1948 ± 17 ha), marsh gain (1085 ± 7 ha), and stable non-marsh (35216 ± 2947 ha) (Table 2). In addition, a reference dataset of 1600 stratified random points was created, yielding a second set of area estimates for stable marsh (7316 ± 378), marsh loss (1514 ± 11 ha), marsh gain (863 ± 3 ha), and stable non-marsh (35523 ± 1982 ha). The mapping accuracy was 85%, with much higher User’s accuracies for the stable classes (92% and 99% for stable marsh and stable non-marsh, respectively) compared to change classes (60% for marsh loss, 61% for marsh gain). In contrast, the Producer’s accuracy estimates derived from stratified random point samples were similar for all classes (80%, 85%, 91%, and 91% for stable marsh, stable non-marsh, marsh loss, and marsh gain, respectively). We base the following results and discussion on estimates derived from the reference dataset, as it derived from a higher-quality workflow (Olofsson et al., 2014). The overall accuracy for the 1995 and 2015 classifications were 90% and 93%, respectively. For the C-CAP data, overall accuracies were 89% and 78% for 1996 and 2016, respectively, while the overall accuracy for 1996–2016 salt marsh trends based on C-CAP data was 75%. The area of estuarine emergent wetlands in C-CAP data for the study region was 9324 ha in 1996 and 9329 ha in 2016 (Table 3).

Table 2.

Confusion matrix for area estimation. Values are in proportion of total area, except where units (in ha) are indicated.

Reference/Map Stable marsh Stable non-marsh Marsh loss Marsh gain Total Area (ha)
Stable marsh 0.1431 0.0050 0.0050 0.0031 0.1563 6968 ± 258
Stable non-marsh 0.0065 0.7642 0.0026 0.0013 0.7745 35216 ± 2947
Marsh loss 0.0061 0.0081 0.0217 0.0001 0.0361 1948 ± 17
Marsh gain 0.0095 0.0035 0.0000 0.0200 0.0331 1085 ± 7
Total 0.1652 0.7809 0.0293 0.0246
Area (ha) 7316 ± 378 35523 ± 1982 1514 ± 11 863 ± 3
Open in a new tab

Table 3.

Estimates of salt marsh area at Barnegat Bay, including previously published values as well as estimates from this study.

Year Marsh area (ha) Study Data source
1888 14,850 Lathrop and Bognar (2001) NJ Geological Service Map
1972 10,472 Lathrop and Bognar (2001) C-CAP
1984 10,380 Lathrop and Bognar (2001) C-CAP
1995 9940 Lathrop and Bognar (2001) C-CAP
1995 8831 ± 389 This study High-resolution orthophoto
1996 9324 This study C-CAP
2007 9321 BBP (2016) NJDEP Land Use Land Cover
2012 9225 BBP (2016) NJDEP Land Use Land Cover
2015 8179 ± 381 This study High-resolution orthophoto
2016 9329 This study C-CAP
Open in a new tab

3.2. Salt marsh change 1995–2015

The area of salt marsh habitat at Barnegat Bay declined by circa 650 ha from 8831 ± 389 ha in 1995–8179 ± 381 ha in 2015 (Table 3; Fig. 3). This is equivalent to a 7.38% loss over two decades, or an annualized loss rate of 0.37% yr−1 (95% CI: 0.07–0.77% yr−1). In absolute terms, the salt marsh loss rate between 1995 and 2015 was estimated at 32.6 ha yr−1 (95% CI: 71.1 to 5.9 ha yr−1).

Fig. 3.

Fig. 3.

Open in a new tab

Salt marsh extent trends at Barnegat Bay from 1995 to 2015. Stable marsh (green) is the area classified as salt marsh in both 1995 and 2015 imagery; Marsh loss (pink) is the area classified as salt marsh in 1995 and non-marsh in 2015; Marsh gain (turquoise) is the area classified as non-marsh in 1995 and salt marsh in 2015.

3.3. Drivers of salt marsh change

Overall, we found that the most prominent driver of marsh loss was ponding, which accounted for 37% of all salt marsh loss. Open marsh water management (OMWM) excavation and the expansion of OMWMs created pre-1995 represented the second-most prominent cause with 27%, while edge erosion accounted for 20% of salt marsh loss. The gains in salt marsh habitat were mainly attributed to revegetation of previously unvegetated bare ground (46%) or ponds (37%), while upland migration was found to be relatively limited (17% of marsh gains).

These estimates of the relative importance of change drivers in concert with the absolute salt marsh areal trends by change class allow for the estimation of area of marsh loss and gain that is caused by each change driver. For example, we found that pond formation was the most prevalent driver of marsh loss at Barnegat Bay, accounting for 37% of marsh loss (Table S6, Fig. 4). The revegetation of previous ponds similarly accounted for 37% of marsh gain. With our estimates of 1514 ± 11 ha of marsh loss and 863 ± 3 ha of marsh gain, we can approximate the area lost to ponding (560 ± 4 ha) and the area gained by revegetation (320 ± 1 ha), with the difference of 240 ± 5 ha representing the deficit of revegetation compared to new pond formation (Table S6). Similarly, we attributed 27% of marsh loss to the excavation and expansion of OMWM ponds and ditches (Fig. 4), resulting in an estimated loss of 409 ± 3 ha to mosquito control activity. Edge erosion resulted in 20% (303 ± 2 ha) of marsh loss (Fig. 4; Table S6). Lastly, we report that 147 ha of salt marsh gain were due to upland migration (e.g., see Fig. 4).

Fig. 4.

Fig. 4.

Open in a new tab

Examples of salt marsh change from 1995 to 2015 associated with the change drivers ‘pond formation’ (top row), ‘mosquito control’ (second row), ‘edge erosion’ (third row), and ‘upland migration’ (bottom row). The left and middle panels show aerial imagery from 1995 to 2015, respectively. The right panels show the marsh trends classification as estimated via object-based image analysis at the same location.

4. Discussion

4.1. Mapping results and accuracy

Because our reference dataset was derived from a higher-quality classification process and imagery compared to the mapping product, the reference-based area estimates form the base of our discussion. It is apparent that the mapping product overestimates the change classes (marsh loss and marsh gain) at the expense of stable classes (stable marsh and stable non-marsh) (Table 2). Error in the mapping products can be introduced by spatial misalignment of source products, which can inflate estimates of marsh gain or loss in high-resolution classifications (Timm and McGarigal 2012). For example, when small features such as ditches or ponds are not perfectly aligned, the true state of stable non-marsh adjacent to stable marsh would be misclassified as an area of marsh gain abutting an area of marsh loss. Furthermore, in a system where the majority of spatial units are expected to be stable, any stochastic mapping error within those stable classes will lead to an erroneous change detection, except for the unlikely event of the error occurring at the same locale in both input products. For future efforts of high-resolution classification of spatially complex habitats such as salt marshes, we emphasize the need for a rigorous protocol for co-registration of source imagery.

Generally, salt marsh mapping can be complicated by a difficulty to establish clear boundaries of often gradual hydrological change in wetland areas. While the presence of emergent salt marsh vegetation may be used to demark marsh boundaries, their identification can be hindered by high tides or plant senescence in winter. The timing of image acquisition is therefore an important limiting factor to wetland mapping, with ideal conditions at low tide and during the growing season. While Barnegat Bay salt marshes are largely microtidal and the tidal stage at image acquisition was favorable (Table 1), the 1995 orthophotos were acquired in April at the start of the growing season, potentially reducing our ability to detect salt marsh vegetation. An additional source of classification error may stem from sun glare impacting open water areas in some of the 1995 imagery. Although our accuracy assessment suggests that 1995 classifications are highly accurate (90%), the reference dataset used to test accuracy was based on photointerpretation of the same orthophotos in addition to Google Earth imagery. Ideally, additional independent data would have been used to test accuracy, but these were not available for 1995. In contrast, the reference data annotation for the 2015 period was based on 0.3 m px−1 orthophotos and several sets of satellite imagery, increasing our confidence in the accuracy assessment for 2015 classifications. Overall, our mapped and reference estimations of marsh area do not yield significant differences for 1995 (8916 ± 276 ha mapped vs. 8831 ± 389 ha reference) or 2015 (8053 ± 265 ha mapped vs. 8179 ± 381 ha reference).

4.2. Salt marsh change detection

The change detection performed in this study estimates an annual loss rate of 0.37% yr−1 (95% CI: 0.07–0.77% yr−1) for Barnegat Bay salt marshes from 1995 to 2015. This loss rate is equivalent to the rate reported for 1984–1995 by Lathrop and Bognar (2001) and intermediate between the loss rates BBP (2016) reported for 1995–2007 (0.52% yr−1) and 2007–2012 (0.21% yr−1).

Previous studies of Barnegat Bay salt marsh change reported significant losses since marshes were mapped by the New Jersey Geological Survey in the late 19th century. The areal extent of salt marsh habitat in 1888 was approximately 14,850 ha but declined to 10,472 ha in 1975 and 10,380 ha in 1984 (Table 3). Lathrop and Bognar (2001) estimated the salt marsh extent in 1995 using NOAA C-CAP data and reported 9940 ha of salt marsh habitat. For the same year, our analysis based on probability sampling arrives at 8831 ± 389 ha and our marsh trend map classifies 8916 ± 276 ha as salt marsh. Furthermore, C-CAP data from 1996 to 2016 both contain larger areas classified as salt marsh compared to our estimates for 1995 and 2015, respectively (Table 3). Weis et al. (2021) showed that for individual marshes in Barnegat Bay, the lower-resolution C-CAP and NJDEP Land Use Land Cover data fail to capture observed changes in marsh extents. This is at least in part due to the heterogeneous structure of salt marshes, with ponds and channels fragmenting the marsh surface. Only high-resolution imagery can adequately distinguish these features, which result in spectral mixing within pixels of larger size (Aplin 2006; Turpie et al., 2015). It is therefore not surprising that our analysis yields smaller salt marsh extent estimates compared to C-CAP data, which likely over-estimate marsh extent by falsely classifying small-scale non-marsh features as salt marsh vegetation.

This notion is further supported by the lower overall accuracies of C-CAP classifications (89% for 1996 and 78% for 2016) compared to our high-resolution products (90% and 93%, respectively), which in part derives from larger errors of commission (Table S3; S4). Similarly, when considering the detection of trends over our 20-year study period, C-CAP-derived trends for Barnegat Bay were less accurate (75%) than those from our high-resolution products (85%); however, C-CAP data for other locations in the U.S. were found to detect change more accurately (85%, McCombs et al., 2016). This suggests that there may be factors introducing classification error that are characteristic of Barnegat Bay salt marshes, particularly in the more recent classifications. Since we found these marshes to be increasingly fragmented, with excavations and ponds contributing to small-scale heterogeneity of the marsh platform, we propose that salt marsh systems in such condition are not appropriately mapped by C-CAP data. However, we acknowledge that high-resolution coastal land-cover data are scarce on a national level and that C-CAP service critical needs of coastal managers. For example, the mapping extent of C-CAP (national) and the frequency of mapping (every 5–10 years) likely preclude high-effort mapping methods such as that presented herein. Our approach necessitates high-resolution imagery and significant time for the annotation of hundreds to thousands of reference points, both of which may not be available to a national wetland mapping program. Still, where salt marsh trends are being evaluated in face of stressors such as nutrient pollution or SLR that can be expected to manifest through marsh fragmentation, ponding, and other forms of marsh deterioration operating on the meter-scale, we advocate for the use of remote sensing products with a resolution equivalent to the size of features of interest, which might be as small as 1 m px−1 (McCarthy et al., 2015; Thomsen et al., 2022). In terms of spectral resolution, we found the inclusion of a near-infrared band to be helpful for the classification of open water (due to signal attenuation) as well as marsh vegetation. In addition, digital elevation models were stacked with spectral bands in our image analysis workflow, which proved useful for classification due to the elevation-dependence of salt marsh vegetation. Consequently, we believe that coastal wetland monitoring can greatly benefit from the availability of high-resolution, multispectral satellite sensors (Timm and McGarigal 2012; McCarthy et al., 2015; Turpie et al., 2015), unmanned aerial vehicles (Krause et al., 2021; Thomsen et al., 2022), and increasingly powerful data processing techniques like structure-from-motion, OBIA, and machine learning (Pierce, 2015; Krause et al., 2021; Thomsen et al., 2022). The inclusion of these technologies into national wetland mapping programs would provide the coastal management community with better tools to adequately address future threats to coastal resources.

At Barnegat Bay, our analysis suggests that salt marsh habitat remains relatively stable and annual loss rates are not accelerating over the last half a century. While Lathrop and Bognar (2001) estimate a loss rate of 50 ha yr−1 from 1888 to 1975, their work indicates a reduced loss rate of 27.6 ha yr−1 between 1972 and 1995. With the present analysis, we estimate the salt marsh loss rate between 1995 and 2015 to be 32.6 ha yr−1 (95% CI: 71.1 to 5.9 ha yr−1), indicating a relatively stable loss rate over the past 40 years. The higher absolute rates of salt marsh loss pre-1972 were mainly attributed to coastal development and fill of tidal wetlands (Lathrop and Bognar 2001), which became more tightly regulated after passing of the Wetlands Act of 1970 and several local regulations in New Jersey. In absence of further coastal development reducing salt marsh area, other anthropogenic and natural drivers can be expected to determine the fate of salt marshes at Barnegat Bay. These drivers are discussed in detail below.

4.3. Drivers of salt marsh change

In the absence of marsh fill for development since 1995, the only direct anthropogenic driver of marsh change was related to mosquito control activities, including new pond excavations for OMWM and the expansion of old ditches and OMWMs. The other drivers of edge erosion, channel widening, and ponding can be considered ‘natural’ or indirect anthropogenic causes of marsh loss and may be attributed to SLR, nutrient pollution, a combination thereof, or other undefined stressors. In contrast, marsh gain occurred through revegetation of bare ground or ponds, as well as upland migration.

Pond formation in salt marshes can be both a natural cycle without net-vegetation loss (Smith and Pellew 2021) or a sign of salt marsh degradation and loss (Watson et al., 2017). When pond dynamics are cyclical, pond evolution is followed by breaching into tidal channels or ditches with subsequent revegetation on time scales of decades (Smith and Pellew 2021). These conditions are thought to be prevalent in un-modified marshes and where large tidal ranges and sufficient sediment supply allow for the pond-depression to fill (Mariotti 2016). However, microtidal marshes with little sediment supply, such as Barnegat Bay, can be more prone to pond formation that leads to permanent marsh-to-open water conversion (Mariotti 2016). The prevalence of ponds on the marsh surface itself was established as an indicator of marsh trajectory, with a larger ratio of unvegetated-to-vegetated surface area indicative of a less healthy marsh (Ganju et al., 2017; Wasson et al., 2019). Furthermore, pond formation may be associated with nutrient pollution through reduction in soil volume (Darby and Turner 2008) and structural changes in soil properties that lead to elevation loss (Wigand et al., 2014). Here, we derive a net-loss of marsh habitat via pond formation of 240 ± 5 ha over two decades (560 ± 4 ha pond formation - 320 ± 1 ha revegetation), which is in line with estimates of other ditched marsh areas in southern New Jersey (Smith and Pellew 2021). Non-cyclical pond formation that maintains unvegetated marsh platform has the potential to permanently lower marsh surface elevation, eventually resulting in a conversion of salt marsh to mudflat (Mariotti 2016; Watson et al., 2017). Management strategies aimed at countering this trajectory involve the addition of dredged sediment to the marsh platform (Ford et al., 1999) or introducing drainage features to reduce surface water impoundments (Wigand et al., 2017; Weis et al., 2021).

Along the mid-Atlantic U.S. coast, salt marshes have been treated for mosquito control by manipulating marsh hydrology. Almost 90% of salt marshes in this region were subjected to ditch construction that was intended to increase draining, as well as OMWM excavation that provides habitat for fish that prey on and reduce mosquito populations (Elsey-Quirk and Adamowicz 2016). Powell et al. (2020) estimate that more than 7000 OMWM ponds were excavated at Barnegat Bay, potentially impairing salt marsh ecosystem function. In addition, dredge spoils resulting from OMWM excavation activities persist as unvegetated bare patches for many decades (Powell et al., 2020). With our analysis, we estimate that the second-most common driver of salt marsh loss at Barnegat Bay between 1995 and 2015 was related to mosquito control, mainly in the form of new OMWM excavation and expansion of previously existing OMWM ponds and ditches (Fig. 4; 5). Of the 1514 ± 11 ha of marsh loss, we attributed 27%, or 409 ± 3 ha to mosquito control activity. This is a considerably larger area than the area of 308 ha of OMWM estimated by Powell et al. (2020), but our estimate also includes ditches. Although salt marshes in the Barnegat Bay region were ditched well before 1995 (Elsey-Quirk and Adamowicz 2016), some ditches, often less than 1 m wide, can erroneously appear as new marsh loss in our trends analysis because of misclassification and geolocation inaccuracies (e.g., see Fig. 4). For improved mapping on the sub-meter scale, unmanned aerial vehicles showed promise (Krause et al., 2021; Thomsen et al., 2022), but their use is currently not practical for studies on regional scales.

Edge retreat is a process by which the seaward edge of the salt marsh platform loses vegetation, and the sediment erodes, thereby converting salt marsh habitat into mud flat or open water. Edge erosion has been documented in salt marshes globally, and loss rates of several meters per year are not uncommon, such as 1–2 m yr−1 in Venice Lagoon (Day et al., 1998), and 4 m yr−1 in the Wadden Sea (Bakker et al., 1993). For Barnegat Bay, edge erosion rates were recently estimated to be below 1 m yr−1 for most of the bay, but up to 2 m yr−1 in some parts of the bay near the inlets, where higher wave exposure led to increased rates of marsh edge retreat (Leonardi et al., 2016). In addition, the process of edge erosion may be accelerated by bank collapse resulting from nutrient pollution (Deegan et al., 2012). The question whether nutrient pollution contributed to our estimate of 303 ± 2 ha of marsh seaward edge erosion is outside the scope of this study, although increasing eutrophication was reported for Barnegat Bay for our study period (Kennish and Fertig 2012).

One of the strategies for the maintenance of salt marsh habitat in the face of relative SLR and eroding marsh seaward edges is allowing the migration of salt marsh into the upland. This process can occur naturally, but coastal development and the protection of coastal infrastructure can prevent upland conversion into salt marsh (Kirwan et al., 2016; Schieder et al., 2018). Nevertheless, undeveloped upland and abandoned agricultural fields represent potential future salt marsh habitat (Wasson et al., 2013; Gedan and Fernández-Pascual 2019). At Barnegat Bay, approximately 70% of upland shoreline buffer is developed and 36% is bulkheaded (Lathrop and Bognar 2001); however, upland migration was found to counteract marsh loss at three Barnegat Bay salt marsh complexes to some degree (Watson, 2019). We found upland migration to account for a marsh gain of 147 ha at Barnegat Bay from 1995 to 2015, which does not suffice to balance the loss terms discussed above. To allow for substantial salt marsh upland migration, coastal management strategies designating marsh migration corridors are needed (Borchert et al., 2018), but their implementation is complicated by the variety of stakeholders with interest in coastal real estate (van Dolah et al., 2020).

4.4. Conclusions

Overall, we report that despite accelerating relative SLR and purported eutrophication, salt marsh loss rates at Barnegat Bay are relatively stable over the past 50 years. The most prevalent drivers of marsh change are not related to ‘marsh drowning’ or nutrient pollution, which would be expected to manifest as extensive net pond formation, bank collapse and edge retreat. In contrast, we attribute the largest portion of salt marsh loss to management activities related to mosquito control, such as OMWM excavation. We note that mosquito control is an important public health concern and that some excavation sites in degraded salt marsh habitat may have been converted to open water even in absence of OMWM excavation. However, since recent studies report a sediment accretion deficit relative to SLR for Barnegat Bay salt marshes (Haaf et al., 2021), we expect more prevalent future salt marsh losses due to climate change-related processes and recommend reducing management practices that directly convert salt marsh habitat to open water.

Supplementary Material

Supplement1

Acknowledgements

This report is ORD Tracking Number ORD-048304, and it has been reviewed technically by the U.S. EPA’s Office of Research and Development, Center for Environmental Measurement and Monitoring, Atlantic Coastal Environmental Sciences Division. The views expressed in this article are those of the authors and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency (EPA). The EPA does not endorse any commercial products, services, or enterprises. Any use of trade, firm or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. This work was supported by a cooperative award by the U.S. Environmental Protection Agency to the Barnegat Bay Partnership (US EPA Regional Applied Research Effort (RARE) award #2062). The authors have no conflict of interest to declare. This is contribution #1518 from the Coastlines and Oceans Division of the Institute of Environment at Florida International University.

Footnotes

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Ethical statement

The authors declare that all ethical practices have been followed in relation to the development, writing, and publication of the article.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.rsase.2022.100910.

Data availability

The link to our data is included in the manuscript under ‘Data availability’

References

  1. Adam P, 2002. Saltmarshes in a time of change. Environ. Conserv 29, 39–61. 10.1017/S0376892902000048. [DOI] [Google Scholar]
  2. Aplin P, 2006. On scales and dynamics in observing the environment. Int. J. Rem. Sens 27, 2123–2140. 10.1080/01431160500396477. [DOI] [Google Scholar]
  3. Bakker JP, de Leeuw J, Dijkema KS, Leendertse PC, Prins HH, Rozema J, 1993. Salt marshes along the coast of The Netherlands. Hydrobiologia 265, 73. 10.1007/BF00007263. [DOI] [Google Scholar]
  4. Barbier EB, Hacker SD, Kennedy C, Koch EW, Stier AC, Silliman BR, 2011. The value of estuarine and coastal ecosystem services. Ecol. Monogr 81 (2),169–193. 10.1890/10-1510.1. [DOI] [Google Scholar]
  5. BBP [Barnegat Bay Partnership], 2016. State of the Bay Report 2016. Barnegat Bay Partnership. Ocean County Community College, Toms River, NJ. https://www.barnegatbaypartnership.org/wp-content/uploads/2017/08/BBP_State-of-the-Bay-book-2016_forWeb.pdf. [Google Scholar]
  6. Blum MD, Roberts HH, 2009. Drowning of the Mississippi Delta due to insufficient sediment supply and global sea-level rise. Nat. Geosci 2, 488–491. 10.1038/ngeo553. [DOI] [Google Scholar]
  7. Borchert S, Osland M, Enwright N, Griffith K, 2018. Coastal wetland adaptation to sea level rise: quantifying potential for landward migration and coastal squeeze. J. Appl. Ecol 55, 2876–2887. 10.1111/1365-2664.13169. [DOI] [Google Scholar]
  8. Cameron Engineering and Associates, 2015. Long Island Tidal Wetlands Trends Analysis. Prepared for the New England Interstate Water Pollution Control Comission. Available at. http://www.dec.ny.gov/docs/fish_marine_pdf/bmrwetlandstrends1.pdf.
  9. Dahl TE, Stedman S, 2011. Status and Trends of Wetlands in the Conterminous United States 2004 to 2009. U.S. Department of the Interior, U.S. Fish and Wildlife Service, Fisheries and Habitat Conservation, Washington, DC, USA. [Google Scholar]
  10. Davidson NC, Fluet-Chouinard E, Finlayson CM, 2018. Global extent and distribution of wetlands: trends and issues. Mar. Freshw. Res 69, 620. 10.1071/MF17019. [DOI] [Google Scholar]
  11. Darby FE, Turner RE, 2008. Below- and aboveground Spartina alterniflora production in a Louisiana salt marsh. Estuar. Coast 31, 223–231. 10.1007/s12237-007-9014-7. [DOI] [Google Scholar]
  12. Day JW, Francesco S, Andrea R, Are D, 1998. Rapid deterioration of a salt marsh in Venice lagoon, Italy. J. Coast Res 14 (2), 583–590. [Available at. http://www.jstor.org/stable/4298811.]. [Google Scholar]
  13. Deegan LA, Hughes JE, Rountree RA, 2002. Salt marsh ecosystems support of marine transient species. In: Weinstein MP, Krieger DA (Eds.), Concepts and Controversies in Tidal Marsh Ecology. Springer Netherlands, Heidelberg, pp. 333–365. 10.1007/0-306-47534-0. [DOI] [Google Scholar]
  14. Deegan LA, Johnson DS, Warren RS, Peterson BJ, Fleeger JW, Fagherazzi S, Wollheim WM, 2012. Coastal eutrophication as a driver of salt marsh loss. Nature 490, 388–392. 10.1038/nature11533. [DOI] [PubMed] [Google Scholar]
  15. Duarte CM, Middelburg JJ, Caraco N, 2004. Major role of marine vegetation on the oceanic carbon cycle. Biogeosci. Discuss 1 (1), 659–679. 10.5194/bg-2-1-2005. [DOI] [Google Scholar]
  16. Elsey-Quirk T, Adamowicz SC, 2016. Influence of physical manipulations on short term salt marsh morphodynamics: Examples from the north and mid-atlantic coast, USA. Estuar. Coast 39, 423–439. 10.1007/s12237-015-0013-9. [DOI] [Google Scholar]
  17. Ford MA, Cahoon DR, Lynch JC, 1999. Restoring marsh elevation in a rapidly subsiding salt marsh by thin-layer deposition of dredged material. Ecol. Eng 12, 189–205. 10.1016/S0925-8574(98)00061-5. [DOI] [Google Scholar]
  18. Ganju NK, Defne Z, Kirwan ML, Fagherazzi S, D’Alpaos A, Carniello L, 2017. Spatially integrative metrics reveal hidden vulnerability of microtidal salt marshes. Nat. Commun 8, 14156. 10.1038/ncomms14156. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Gedan KB, Fernández-Pascual E, 2019. Salt marsh migration into salinized agricultural fields: a novel assembly of plant communities. J. Veg. Sci 30 (5), 1007–1016. 10.1111/jvs.12774. [DOI] [Google Scholar]
  20. Haaf L, Burke Watson E, Elsey-Quirk T, Raper K, Padeletti A, Maxwell-Doyle M, Kreeger D, Velinsky DJ, 2021. Sediment accumulation, elevation change, and the vulnerability of tidal marshes in the Delaware estuary and Barnegat bay to accelerated sea level rise. Estuar. Coast 10.1007/s12237-021-00972-9. [DOI] [Google Scholar]
  21. Kearney MS, Rogers AS, Townshend JR, Rizzo E, Stutzer D, Stevenson JC, Sundberg K, 2002. Landsat imagery shows decline of coastal marshes in Chesapeake and Delaware Bays. Eos 83, 173–178. 10.1029/2002EO000112. [DOI] [Google Scholar]
  22. Kennish MJ, 2001b. Physical description of the Barnegat bay-little egg harbor estuarine system. J. Coast Res (32), 13–27. [Available at. http://www.jstor.org/stable/25736225.]. [Google Scholar]
  23. Kennish MJ, 2001a. Coastal salt marsh systems in the U.S.: a review of anthropogenic impacts. J. Coast Res 17 (3), 731–748. [Available at. http://www.jstor.org/stable/4300224.]. [Google Scholar]
  24. Kennish MJ, Fertig B, 2012. Application and assessment of a nutrient pollution indicator using eelgrass (Zostera marina L.) in Barnegat Bay–Little Egg Harbor estuary, New Jersey. Aquat. Bot 96, 23–30. 10.1016/j.aquabot.2011.09.005. [DOI] [Google Scholar]
  25. Kirwan ML, Murray AB, Boyd WS, 2008. Temporary vegetation disturbance as an explanation for permanent loss of tidal wetlands. Geophys. Res. Lett 35 (5). 10.1029/2007GL032681. [DOI] [Google Scholar]
  26. Kirwan ML, Walters DC, Reay WG, Carr JA, 2016. Sea level driven marsh expansion in a coupled model of marsh erosion and migration: sea Level Driven Marsh Expansion. Geophys. Res. Lett 43, 4366–4373. 10.1002/2016GL068507. [DOI] [Google Scholar]
  27. Koch EW, Barbier EB, Silliman BR, Reed DJ, Perillo GME, Hacker SD, Granek EF, et al. , 2009. Non-linearity in ecosystem services: temporal and spatial variability in coastal protection. Front. Ecol. Environ 7 (1), 29–37. 10.1890/080126. [DOI] [Google Scholar]
  28. Krause JR, Burke Watson E, Maher N, 2020. Are tidal salt marshes exposed to nutrient pollution more vulnerable to sea level rise? Wetlands. 10.1007/s13157-019-01254-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Krause JR, Hinojosa-Corona A, Gray AB, Burke Watson E, 2021. Emerging sensor platforms allow for seagrass extent mapping in a turbid estuary and from the meadow to ecosystem scale. Rem. Sens 13, 3681. 10.3390/rs13183681. [DOI] [Google Scholar]
  30. Lathrop RG, Bognar JA, 2001. Habitat loss and alteration in the Barnegat bay region. J. Coast Res (32), 18. [Available at. http://www.jstor.org/stable/25736235.]. [Google Scholar]
  31. Leonardi N, Defne Z, Ganju NK, Fagherazzi S, 2016. Salt marsh erosion rates and boundary features in a shallow bay: indicators salt marsh erosion New Jersey. J. Geophys. Res.: Earth Surf 121 (10), 1861–1875. 10.1002/2016JF003975. [DOI] [Google Scholar]
  32. Lotze HK, Lenihan HS, Bourque BJ, Bradbury RH, Cooke RG, Kay MC, Kidwell SM, Kirby MX, Peterson CH, Jackson JBC, 2006. Depletion degradation, and recovery potential of estuaries and coastal seas. Science 312, 1806–1809. 10.1126/science.1128035. [DOI] [PubMed] [Google Scholar]
  33. Macreadie PI, Costa MDP, Atwood TB, Friess DA, Kelleway JJ, Kennedy H, Lovelock CE, Serrano O, Duarte CM, 2021. Blue carbon as a natural climate solution. Nat. Rev. Earth Environ 10.1038/s43017-021-00224-1. [DOI] [Google Scholar]
  34. Mariotti G, 2016. Revisiting salt marsh resilience to sea level rise: are ponds responsible for permanent land loss?: salt marsh ponds. J. Geophys. Res.: Earth Surf 121 (7), 1391–1407. 10.1002/2016JF003900. [DOI] [Google Scholar]
  35. McCarthy MJ, Merton EJ, Muller-Karger FE, 2015. Improved coastal wetland mapping using very-high 2-meter spatial resolution imagery. Int. J. Appl. Earth Obs. Geoinf 40, 11–18. 10.1016/j.jag.2015.03.011. [DOI] [Google Scholar]
  36. McCombs JW, Herold ND, Burkhalter SG, Robinson CJ, 2016. Accuracy assessment of NOAA coastal change analysis program 2006–2010 land cover and land cover change data. Photogramm. Eng 82, 711–718. 10.14358/PERS.82.9.711. [DOI] [Google Scholar]
  37. Olofsson P, Foody GM, Herold M, Stehman SV, Woodcock CE, Wulder MA, 2014. Good practices for estimating area and assessing accuracy of land change. Remote Sens. Environ 148, 42–57. 10.1016/j.rse.2014.02.015. [DOI] [Google Scholar]
  38. Partnership for the Delaware Estuary [PDE], 2012. Technical report for the delware estuary and basin. PDE Report 12–01. 255 pages. www.delawareestuary.org/science_programs_state_of_the_estuary.asp.
  39. Pierce K, 2015. Accuracy optimization for high resolution object-based change detection: an example mapping regional urbanization with 1-m aerial imagery. Rem. Sens 7, 12654–12679. 10.3390/rs71012654. [DOI] [Google Scholar]
  40. Powell EB, Krause JR, Martin RM, Burke Watson E, 2020. Pond excavation reduces coastal wetland carbon dioxide assimilation. J. Geophys. Res.: Biogeosciences 125 (2), 1–19. 10.1029/2019JG005187. [DOI] [Google Scholar]
  41. Schieder NW, Walters DC, Kirwan ML, 2018. Massive upland to wetland conversion compensated for historical marsh loss in Chesapeake bay, USA. Estuar. Coast 41, 940–951. 10.1007/s12237-017-0336-9. [DOI] [Google Scholar]
  42. Smith SM, 2009. Multi-decadal changes in salt marshes of Cape Cod, MA: photographic analyses of vegetation loss, species shifts, and geomorphic change. Northeast. Nat 16, 183–208. 10.1656/045.016.0203. [DOI] [Google Scholar]
  43. Smith JAM, Pellew M, 2021. Pond dynamics yield minimal net loss of vegetation cover across an unditched salt marsh landscape. Estuar. Coast 44 (6), 1534–1546. 10.1007/s12237-020-00882-2. [DOI] [Google Scholar]
  44. Thomsen AS, Krause JR, Appiano M, Tanner KE, Endris C, Haskins J, Burke Watson E, Woolfolk A, Fountain MC, Wasson K, 2022. Monitoring Vegetation Dynamics at a Tidal Marsh Restoration Site: Integrating Field Methods, Remote Sensing and Modeling. Estuaries and Coasts. 10.1007/s12237-021-00977-4. [DOI] [Google Scholar]
  45. Timm BC, McGarigal K, 2012. Fine-scale remotely-sensed cover mapping of coastal dune and salt marsh ecosystems at Cape Cod National Seashore using Random Forests. Remote Sens. Environ 127, 106–117. 10.1016/j.rse.2012.08.033. [DOI] [Google Scholar]
  46. Turner RE, Howes BL, Teal JM, Milan CS, Swenson EM, Tonerb DDG, 2009. Salt marshes and eutrophication: an unsustainable outcome. Limnol. Oceanogr 54 (5), 1634–1642. 10.4319/lo.2009.54.5.1634. [DOI] [Google Scholar]
  47. Turpie KR, Klemas VV, Byrd K, Kelly M, Jo Y-H, 2015. Prospective HyspIRI global observations of tidal wetlands. Remote Sens. Environ 167, 206–217. 10.1016/j.rse.2015.05.008. [DOI] [Google Scholar]
  48. Velinsky DJ, Paudel, Quirk T, Piehler M, Smyth A, 2017. Salt marsh denitrification provides a significant nitrogen sink in Barnegat bay, New Jersey. J. Coast Res 78 (May), 70–78. 10.2112/si78-007.1. [DOI] [Google Scholar]
  49. Van Dolah ER, Miller Hesed CD, Paolisso MJ, 2020. Marsh migration, climate change, and coastal resilience: human dimensions considerations for a fair path forward. Wetlands 40, 1751–1764. [Google Scholar]
  50. Wasson K, Woolfolk A, Fresquez C, 2013. Ecotones as indicators of changing environmental conditions: rapid migration of salt marsh–upland boundaries. Estuar. Coast 36 (3), 654–664. 10.1007/s12237-013-9601-8. [DOI] [Google Scholar]
  51. Wasson K, Jeppesen R, Endris C, Perry DC, Woolfolk A, Beheshti K, Haskins J, et al. , 2017. Eutrophication decreases salt marsh resilience through proliferation of algal mats. Biol. Conserv 212, 1–11. 10.1016/j.biocon.2017.05.019. [DOI] [Google Scholar]
  52. Wasson K, Ganju NK, Defne Z, Endris C, Elsey-Quirk T, Thorne KM, Freeman CM, Guntenspergen G, Nowacki DJ, Raposa KB, 2019. Understanding tidal marsh trajectories: evaluation of multiple indicators of marsh persistence. Environ. Res. Lett 14 (12), 124073. 10.1088/1748-9326/ab5a94. [DOI] [Google Scholar]
  53. Watson EB, Oczkowski AJ, Wigand C, Hanson AR, Davey EW, Crosby SC, Johnson RL, Andrews HM, 2014. Nutrient enrichment and precipitation changes do not enhance resiliency of salt marshes to sea level rise in the Northeastern U.S. Climatic Change 125, 501–509. 10.1007/s10584-014-1189-x. [DOI] [Google Scholar]
  54. Watson EB, Wigand C, Davey EW, Andrews HM, Bishop J, Raposa KB, 2017. Wetland loss patterns and inundation-productivity relationships prognosticate widespread salt marsh loss for southern new England. Estuar. Coast 40 (3), 662–681. 10.1007/s12237-016-0069-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Watson EB, 2019. A model-data synthesis of the status and trends of New Jersey’s coastal wetlands for sea level rise planning. Final report to NJ sea grant. Project number: 6610–0005. 82 Pages. https://macwa.org/assets/img/com/Final_Report.pdf.
  56. Weis JS, Burke Watson E, Ravit B, Harman C, Yepsen M, 2021. The status and future of tidal marshes in New Jersey faced with sea level rise. Anthropocene Coast. 4, 168–192. 10.1139/anc-2020-0020. [DOI] [Google Scholar]
  57. Wigand C, Ardito T, Chaffee C, Ferguson W, Paton S, Raposa K, Vandemoer C, Watson E, 2017. A climate change adaptation strategy for management of coastal marsh systems. Estuar. Coast 40, 682–693. 10.1007/s12237-015-0003-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Wigand C, Roman CT, Davey E, Stolt M, Johnson R, Hanson A, Rafferty P, et al. , 2014. Below the disappearing marshes of an urban estuary: historic nitrogen trends and soil structure. Ecol. Appl 24 (4), 633–649. 10.1890/13-0594.1. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplement1
NIHMS1895626-supplement-Supplement1.docx (545.3KB, docx)

Data Availability Statement

The link to our data is included in the manuscript under ‘Data availability’

RESOURCES