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, 8,830 ± 390 ha were covered with marsh vegetation, while in 2015 only 8,180 ± 380 ha of salt marsh habitat remained.

Data for the classification of coastal land-cover of Barnegat Bay were downloaded from the USGS Earth Resources Observation and Science (EROS) Center via the USGS Earth Explorer website. The spatial data comprised high-resolution orthoimages acquired in April 1995 (color-infrared, 3 bands, 1 m resolution) and April 2015 (4 bands, red, green, blue, near-infrared, 0.3 m resolution), as well as a bathymetric model from 2000 (10 m resolution) and a digital elevation model (DEM) from 2015 (1 m resolution). 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). Aerial imagery orthomosaics were clipped to the same extent and fused 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 imported into eCognition Developer 9 (Trimble, Sunnyvale, CA, USA) for object-based image analysis. After spectral-difference and multi-resolution segmentation, 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 to 5, maximum difference, hue, or brightness via hierarchical decision trees. Obvious classification errors were manually corrected before classification results were exported as raster files. 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.

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. Labeling of the reference dataset used higher resolution imagery, where available. In addition, drivers of changes in Barnegat Bay salt marsh extent were annotated 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.

Files are provided as raster in .tiff format (8 bit).

Classes are defined as follows:

1995.tif, 2015.tif: 1-Open Water; 2-Upland; 3-Salt Marsh; 4-Unvegetated

Trend.tif: 1-Stable non-marsh; 2-Stable marsh; 3-Marsh gain; 4-Marsh loss

Ground-control points are provided in .csv format.

Classes are defined as follows:

Control_points.csv: M-Marsh; U-Unvegetated