Data for: Plant thresholds and community composition of coastal marsh-forest ecotones in the US Northeast

Sea level rise is causing coastal salt marshes to migrate upslope into coastal forests and other terrestrial ecosystems.  However, the factors that control marsh migration rates are not well understood, particularly in the US Northeast, where this phenomenon has received little attention. To determine the relationship between environmental variables and plant species composition in marsh migration zones, we examined plant coverage and environmental data for three sites experiencing marsh upslope migration in New Jersey, New York, and Massachusetts that varied in slope from 1-3%. We found that only 10% of the variation in plant community composition was explained by inundation time, while models containing multiple predictor variables, including edaphic variables, explained much greater levels of variance. Random forest models predicting native halophyte presence /had accuracies of 69 – 84%, with salinity, flooding duration, and light availability as key predictors. The most accurate model for mature tree presence (84% accuracy) highlighted salinity and flooding time as the most important variables. Threshold Indicator Taxa ANalysis (TITAN) identified plant community changepoints at soil salinities of 0.8 and 7.6 PSU, reflecting the lower and upper boundaries of the marsh-forest ecotone. Treelines varied in elevation by site, suggesting that the amount of tidal flooding that trees can withstand varies. These results highlight the importance of multiple variables interacting to determine species distributions and community composition in marsh-forest ecotones.

Methods
Site Descriptions

Three sites in the northeast exhibiting forest retreat along the marsh-forest ecotone were chosen for study, including Waquoit Bay (WB), Massachusetts (41.5562°, -70.5057°), part of the Waquoit Bay National Estuarine Research Reserve. The site has a tidal range of 0.61 m and an ecotone-upland slope of 3.45% ± 0.46. The second site, Pine Neck, is part of Pine Neck Preserve in East Quogue, NY (40.8422°, -72.5644°) and is managed by the Nature Conservancy. This site has a tidal range of 0.78 m and an ecotone-upland slope of 2.12% ± 0.58. The southernmost site selected was Egg Harbor, part of the Tuckahoe Wildlife Management Area (39.3259°, -74.6502°) in southern New Jersey. This site has a tidal range of 1.14 m and has the lowest slope of the three, 1.80% ± 0.64. Soils along the marsh-upland border transitioned from muck (classified as Freetown and Swansea muck) or Transquaking peat at low elevations to sandy soils (Carver coarse sand, Deerfield loamy sand, Plymouth loamy sand, Hammonton sandy loam) at higher elevations. Sandy soils are ubiquitous along the coastal plain and originated from glacial outwash in NY and MA and marine deposition in NJ.

Field Sampling

At each site, sampling points were selected to form a grid with 26.5 m spacing spanning the high marsh and forest edge. We navigated to sampling points using a Garmin GPSMap 64st (Garmin Ltd., Olathe, KA). At each sampling point, a 0.5 m2 quadrat was haphazardly dropped, and percent cover of understory plant taxa (including saplings) was estimated within the plot. The presence of live mature trees was noted for a 5 x 5m area centered around the plot. A 100-cc soil ring sampler was used to collect a 5cm-deep soil core from each plot. Cores were sealed in plastic bags and frozen after returning from the field. Light availability for each sampling point was measured for five minutes at a height of 3 m using a Photosynthetically Active Radiation (PAR) sensor (Odyssey Light Meters, Dataflow Systems Ltd., Christchurch, NZ) relative to a sensor placed in an unobstructed canopy to correct for cloud cover. Depth to groundwater was recorded as either: (a) the maximum depth where water could be obtained using a pushpoint porewater sampler designed to sample saturated groundwaters (MHE products, East Tawas, MI, USA) or (b) depth to water table measured in a shallow borehole. All groundwater depths greater than 84 cm were recorded as >84 cm.

Lab Analyses

In the lab, edaphic conditions were characterized for salinity, redox, moisture, bulk density, and organic matter content. Redox was measured from the center of each core using a Sper Scientific benchtop meter (Sper Scientific, Ltd., Scottsdale, AZ). Soil cores were then dried at 70° C until the mass no longer decreased after consecutive measurements. Gravimetric moisture content was measured as the percentage mass lost during drying. Soil bulk density was calculated by dividing the dry mass by the volume of the core. Soil organic matter content was measured using Loss on Ignition (LOI), where samples were combusted at 550° C for four hours, and organic matter content was recorded as the percentage mass that was lost through combustion. A volume of deionized water equal to 5x the soil mass was added to 3-5 cc of dry soil, vortexed for 10 seconds, held overnight, centrifuged for 5 minutes at 2500 RPM, and salinity was measured on the supernatant using a conductivity/temperature probe attached to a YSI Professional Plus meter (Xylem Inc., Yellow Springs, OH). Salinity was measured on duplicate soil subsamples for three plots per site. Salinity is reported as the 5:1 slurry of deionized water and dry soil. The average water:soil ratio of the samples we collected was similar at 3:1.

*Geospatial Measurements *

Elevations of the sampling points were extracted from bare-earth Digital Elevation Models (DEMs) derived from LiDAR surveys. A comparison with static GPS measures (10 per site spanning the marsh to the forest) found an average difference between LiDAR and GPS-measured elevations of 0.02 m. To compare elevations of tree boundaries among sites, the lower tree boundary was digitized manually from high-resolution (0.6 m) aerial imagery acquired in 2018/9 from the National Agriculture Imagery Program (NAIP). Living trees that were within 15 m from their nearest neighbor were included as part of the boundary, while trees >15 m from nearest neighbors were excluded. Boundaries were digitized 3 times so that digitizer error could be estimated as the standard deviation of the mean elevation from each attempt. The upper boundary of salt-tolerant plants was delineated through field surveys using a Garmin 64st (Accuracy ± 3 m) (Garmin Ltd., Olathe, KS), where the uppermost boundary of salt-tolerant plants was delineated. The uppermost salt-tolerant plants were most commonly Baccharis halimifolia or Phragmites australis. The elevation distribution of this field-derived demarcation was estimated using bare-earth DEMs in ArcGIS Pro (version 3.2.1, ESRI, Redlands, CA, USA). Because tidal range varied between sites, elevations standardized for tidal range (z*) were generated as the elevation in meters relative to mean sea level, divided by the tidal range (z* = (Z-MSL)/(MHHW-MSL)), where z = mean elevation of the boundary, and MSL and MHHW = tidal datums acquired from VDatum.