Landsat 5 and 7 annual NDVI composites from 1989 to 2013 were acquired from the Google Earth Engine website.\u00a0Handheld GPS units (Garmin etrex 20) were used to outline and ground truth the mangrove die-off area in June 2014. Historical hurricane tracks were acquired from the National Oceanic and Atmospheric Administration (NOAA) Digital Coasts website.
\nDie-off GPS outline.CSV: This file consists of GPS coordinates collected by a handheld GPS of the die-off area. This was used to create a shapefile of the die-off area in GIS.
These data were compiled to better determine when the mangrove die-off began. NDVI values of 0.2 or greater correspond to live mangrove cover. These data suggest the mangrove die-off may have started in 2008 and was exacerbated in following years with some recovery in 2013. Live area was calculated by determining the area of the die-off region with NDVI values greater than 0.2.
For Landsat NDVI composites: Years with severe cloud cover were removed from the dataset resulting in the exclusion of 1990 and 2003. Images were transformed to Normalized Difference Vegetation index (NDVI) following Chander, et al. [22] by USGS. We used coordinates from handheld GPS units to create an outline of the die-off area as a shapefile using ArcMap 10.3.1. This shapefile was checked against a basemap of the Caribbean produced by the U.S Geological Survey (USGS) and high-resolution imagery from WorldView-2 (http://www.satimagingcorp.com/satellite-sensors/worldview-2/). Annual NDVI composites from each year were clipped to the shapefile in ArcMap 10.3.1. NDVI was reclassified into five categories (Table I). Area of each reclassified NDVI category was calculated for every year in the die-off region using R (version 3.4.1). NDVI values categorized as yellow and pale yellow (0.2-0.5, 0.5-0.7, respectively) were considered as live mangrove. Historical hurricane tracks were filtered such that only storms that directly passed over Abaco Island between 1984 and 2013 were included in our study (Table II). We constructed a time series of live mangrove vegetation cover and of maximum annual NDVI from 1989-2013 in R (version 3.4.1). Maximum annual NDVI was chosen to ensure minor cloud cover did not affect the overall values and reflects the maximum NDVI value over the entire die-off area. We then superimposed hurricane events on the time series to determine if a pattern between storm events and changes in mangrove cover and maximum NDVI were evident.
\nBCO-DMO Data Processing Notes:
\n-Reformatted column names to comply with BCO-DMO standards
\n-Replaced blanks with nd