Supervised LULC Classifications
To analyze the current land use and land cover (LULC) for the six SEAS properties, we conducted supervised LULC classifications for each property. We first acquired GIS data and imagery files, including property boundary shapefiles, high resolution aerial imagery from Nearmap, and 2018 color infrared imagery from the USDA’s National Agriculture Imagery Program (NAIP). Using the NAIP imagery and ArcGIS Pro software, training samples were created for the seven main land cover types from the National Land Cover Database 2011 (NLCD2011) - developed, deciduous forest, evergreen (coniferous) forest, herbaceous, planted/cultivated, wetlands, and water. Nearmap (natural color, leaf-on) and the ESRI Basemap: World Imagery (natural color, leaf-off) were also used to help interpret the different land cover types.
Once a sufficient number of training sample polygons were created to represent the full spectral range within each class (e.g. darkest to lightest areas of water), a supervised classification was run using the Support Vector Machine method to create the final LULC outputs. To calculate the area (in hectares) of each of the classified land cover types, the LULC raster was transformed into a polygon feature class. A new field was then created in the attribute table to calculate the areas and summarize the results.
Combining LULC with National Wetlands Inventory data
Unlike forested land cover, supervised LULC classifications for wetlands do not have a high accuracy rate as there is a high probability that forested wetlands will be classified as forest
and emergent wetlands will be classified as herbaceous. Therefore, to get a more accurate assessment of the wetlands land cover area for each of the properties, the supervised LULC
classifications were rerun without the training samples for wetlands. The latest data from the National Wetlands Inventory (USFWS, 2018) was then downloaded and clipped to the individual property boundary polygons.
After converting the supervised LULC raster files to polygons, the wetlands polygons were erased from the LULC polygons. In the LULC attribute table, a new field was created to calculate the area in hectares for each cover type and summarize the results. The same was done for the wetlands polygons to find the area of each wetland type (freshwater forested/shrub and freshwater emergent).