Ensure you explain WHY your selected bands provide maximum contrast, i.e. what controls absorption and scattering in each band and are these biophysical properties present in the features in the image. Spectral signatures could also be shown for key land-cover features to explain why your band combination works.
Which band of combination and form of contrast stretch provides maximum contrast between ALL land cover types in an image? Use histograms or reflectance signatures to explain band selection and how each stretch was implemented, specifically which range(s) of DN were enhanced and why. Include reference and key points from relevant supporting literature on each of the contrast stretching algorithms.
Provide an explanation of which environmental features the NDVI and IRON OXIDE image appear sensitive to. Use vegetation and soil signatures showing how they absorb, transmit and scatter light to explain how the indices work.
Provide suggestions of how to improve the image classification to improve the discrimination between land cover classes.
Explain how the image classification processing operation is used to transform the Landsat OLI image into a map showing the spatial distribution of different vegetation types on the land areas of the image. Ensure this covers everything from class selection, training, classification and labelling (possibly accuracy assessment) and reference appropriate literature to support it.
Compare and contrast the following two types of image classification operations in terms of their ability to produce an accurate land-cover map: supervised classification and unsupervised classification. Make sure that the all of the stages in the supervised and unsupervised process are presented in sequence, from selecting classes to map, generating training sites……. to production of an output map with information class. You can list the supervised and unsupervised approach at each stage. Also include text to highlight the applications in which you would use supervised (i.e. you have field reference data and know the area) or unsupervised (you do not have reference data and don’t know the area).
