The data came from MassGIS, the state GIS agency of Massachusetts (http://www.mass.gov/mgis/). The original resolution is 0.3 meter. We downsized it into 0.5 meter.
The salt marshes (both Spartina alterniflora and Spartina patens) are very spectrally similar with grass. To avoid misclassification led by this problem, a wetland mask is used to separate marsh area from other land cover types. This mask map is downloaded from MassGIS. Because the wetland map has questionable accuracy, around 20 hours was spent first to manually refine this mask using our latest 2013 fine-resolution image. After the ‘wetland’ part and ‘no-wetland’ part were separated from the original image, object-based classification was performed for each part. For the ‘wetland’ part, we defined the classification system as four categories, i.e. ‘Water’, ‘Tidal flats/soil’, ‘Spartina alterniflora’ and ‘Spartina patens’. For the ‘non-wetland’ part, we defined five categories, i.e. ‘water’, ‘Tidal flats/soil’, ‘Trees’, ‘Grass’ and ‘Impervious’. This approach enables separately classifying ‘grass’ and two marsh species without confusing these two.

Object-based classification was performed using the module ‘Example-based feature selection’ in the software Exelis Visual Information Solutions 5.0 (ENVI 5.0). The following is the detailed description for object-based classification. First, the image was segmented into many homogeneous ‘objects’ by using a segmentation technique; the best parameters for segmentation were determined by a trial-and-error procedure. Then, we selected ‘blue band’, ‘green band’, ‘red band’, ‘NIR’ and ‘NDVI’ as five properties for each object, and performed supervised classification using k-NN algorithm. For more details, we refer readers to the help document in ENVI software.

We applied the procedure described above to every tile for each image. After classification, approximately 30 additional hours were taken to perform manual edition for each map in order to guarantee the thematic accuracy in our final maps.

Information relevant to the GIS data encoding

This is a seven-category land-cover map of the Plum Island Sound estuary, Massachusetts. The seven categories are: water, tidal flats and soils, Spartina alterniflora, Spartina patens, trees, grass, impervious surface.

Geographic Coordinate Reference: GCS_North_American_1983
Projection: North American 1983 UTM Zone 19M
Reference Ellipsoid: Name: GRS 1980 Semi Axis: 6378137.0
Meridian: Greenwich
Number of bands : 1
Cell Geometry : area

Additional processing information is available via Sussman2017_FinalReport_SuYe.pdf which can be found in the  LC_2013_PlumIslandEstuary_raster_V1.zip Data Source file for this data set.