Data Sources

Nineteen Landsat 5 Thematic Mapper (TM) and four Landsat 8 Operational Land Imager (OLI) image acquisition dates were selected for analysis (table 1). Landsat standard data products, processed using the Level 1 Product Generation System (LPGS), were downloaded from the USGS Earth Resources Observation and Science (EROS) Center data archives. Image acquisition dates correspond to 5-year intervals beginning in 1984 supplemented with additional images from intervening years for which high-resolution aerial photography was also analyzed (Guy, 2015). If available, spring (April preferred) and late fall (November preferred) image acquisition dates were used for each analysis year. Monthly water levels measured at the long-term tide gage at Ocean City Inlet, Maryland show the smallest average seasonal deviation from mean sea level during these months (Zervas, 2009); in addition, analysis of spring and fall scenes provides information on seasonal variation in vegetation response. The spring and fall images were supplemented with additional scenes corresponding to aerial photography acquisition dates (Guy, 2015). A gap in imagery dates from August 31, 2011, to April 14, 2013, corresponding to the decommissioning of the Landsat 5 TM mission and the first data available from the recently launched Landsat 8 OLI mission. Data from the Landsat 7 Enhanced Thematic Mapper Plus (ETM+) satellite are available for this time period; however, data gaps exist over the study area in all Landsat 7 images acquired since May 2003 due to the failure of the scan line corrector (U.S. Geological Survey, 2012), limiting the use of these data in quantitative land-change analyses.

Compared to the high-resolution aerial photography (1-foot to 2-m pixel resolution) analyzed by Guy (2015), the location of features such as digitized shoreline or sand-line positions derived from satellite-based Landsat imagery (30-m pixel resolution for reflective bands used in land-cover classification) may be less precise; however, Landsat imagery offers several advantages over aerial photography datasets. Each Landsat scene covers an area approximately 175 km by 183 km, minimizing the number of images necessary to encompass large study areas and making Landsat data ideal for regional- or landscape-scale analyses. Landsat images are acquired every 16 days, enabling trends analyses at varying scales (for example, seasonal, annual, and decadal), whereas aerial photography missions are flown at irregular intervals. Because Landsat images include multiple infrared bands in addition to the visible red, blue, and green spectrum, various classification algorithms can be used to differentiate between land-cover or habitat types, and shoreline positions and land-cover boundaries can easily be extracted from the classed imagery; these features are commonly hand-digitized from aerial photographs.