Wetland classification model developed with remotely sensed imagery obtained from the Sentinel-1 and -2 satellites and digitized species distribution maps for southwest Florida, coastal Gulf of Mexico, from 2010 to 2018 (NCEI Accession 0243071)

A hierarchical vegetation classification model (10 m resolution) was developed for southwest Florida wetlands using a fusion of multispectral and synthetic aperture radar (SAR) remotely sensed imagery. Sentinel-1 and 2 imagery were obtained from Dec 2015-Sept 2017, split into wet and dry seasons, and processed for a range of vegetation and multi-temporal indices for a total of 26 predictor layers. Training datasets included polygons developed from field surveys and high resolution imagery collected from 2010 - 2018. The domain was first split into estuarine and interior wetlands, then an open water, forest, or grassland model (high level) was developed for each wetland type. Finally, classification model that included species and community-level classes (fine level) was created. Mean overall accuracy was 0.90 and 0.80 for the high and low level models, respectively.