Supervised Classification

We generated supervised classification maps for all our datasets. This involved steps to create vector maps that contained training areas; digitizing training areas of different classes (in our case, five classes); converting the vector training maps into raster maps; generating spectral signature statistics; and finally classifying cell spectral reflectances using the maximum likelihood discriminant analysis classifier which uses the region means and covariance matrices from the spectral signature file, based on regions (groups of image pixels) chosen by us, to determine to which category each cell in the image has the highest probability of belonging” (GRASS GIS Manual). We digitized different features using NDVI files of each year with corresponding false colored composites as reference images during the digitization. The digitized features were assigned to the following five land use and land cover classes— Class 1. Water Bodies; Class 2. Rocky/Cliff Faces/Shadows; Class 3. No Vegetation which includes bare soil; Class 4. Light to Moderate Vegetation representing agricultural land, grasses, shrubs; and Class 5. Dense Vegetation that represents the forest cover.  On average we digitized a minimum of 80 training areas that best represented the different classes for each dataset.