Coastal Change Likelihood in the U.S. Northeast Region: Maine to Virginia - Fabric Dataset

All data were checked for accuracy during processing and an accuracy assessment was conducted on the composite landscape delineation (see associated data report), which suggests that the overall landscape classification is around eighty percent accurate with a Kappa coefficient of 0.77, which is considered ‘in substantial agreement’ with the source aerial imagery. Due to the large number of source datasets that have been combined and merged in this dataset, there are some known inconsistencies, and there may be others unlisted or yet unrealized. The source data often have incomplete extents, gaps, and inconsistencies, which primarily arise from the resolution, quality, density, format, date, source imagery, or spatial processing with other data sources. Some of the known issues are listed here.
Elevation inconsistencies: CONED elevation data were the preferred source of high-resolution seamless bathymetry and topography for the study area. However, CONED data north of Cape Ann, MA are not yet available. Additional DEM information was added to CONED to fill in gaps in northern MA, NH and ME. The offshore domain in ME and NH was approximated due to a lack of high resolution, seamless bathymetric data. The offshore domain is extended to the edge of the source CCAP landcover extent, but does not uniformly represent the –10-meter contour like areas where CONED data are the source elevation information
Landcover inconsistencies: 10 mpp Beta CCAP landcover data were the definitive source of landcover data wherever available. Alternative land use/land cover (LULC) data sources were used for New Jersey, Virginia, and Pennsylvania (along the Delaware River) where 10 mpp CCAP data are not available. The change in source landcover data created some seams along and within state boundaries, as well as some differences in landscape classification. For example, the New Jersey landcover source associates ‘recreational lands’ with ‘Urban’ landscapes, which is different than CCAP. This difference in classification algorithm causes parts of New Jersey to appear more developed compared to other states. Landcover data for the Aberdeen Proving Ground military installation are masked for security purposes and have been given a ‘NoData’ classification, which is the same as aquatic class. A change in imagery date in the MD CCAP data creates an unnatural boundary in the landcover data in part of MD.
Shoreline data inconsistencies: Three landscape classes in this study were derived from vector shoreline data (NOAA Environmental Sensitivity Index (ESI) datasets listed in the source data section). Both polyline and polygon datasets were rasterized and merged with the LULC source data. ESI shoreline classifications have some inconsistencies between the regional and national data products, and several of these can be seen in the landscape classification. For example, differences in the MD, VA and Chesapeake Bay ESI datasets may result in overlapping or misaligned features (such as seawalls). Spatial alignment issues arise from datum adjustments and shoreline source differences in the LULC data and the ESI data. Additionally, the rasterization of vector data can create irregularly or jagged shorelines when converted to pixels. Although the translation of these vector shoreline feature classifications is not perfect, the overall benefit of including the three additional landscape classes from vector sources (rocky, hardened, and tidal flats) in this study far exceeds the limitations of a coastal landscape classification without them.
Shoreline change data inconsistencies and completeness: The shoreline change rates associated with the unconsolidated shore landclass are more extensive for the state of MA than they are for other states in this study. Shoreline change data derived from the National Assessment of Shoreline change (see source listed below) are located along sandy, open-ocean-facing coasts. However, the MA shoreline change data often covers embayments and backbarrier shorelines. This difference in data availability results in more detailed assessment of the shoreline in MA compared to other states.
UVVR data resolution: The data from the Unvegetated to Vegetated Ratio (UVVR) are the only source fabric data that did not exist at 10 mpp or finer resolution. The 30 mpp UVVR product was used in this study to resolve the decision nodes associated with saltmarshes, and this affects the apparent resolution of salt marshes where UVVR was applied.
For a more complete explanation of how source data affect the results of the CCL dataset as a whole, see the discussion section of the associated data report (https://doi.org/10.3133/dr1169)