My GIS MSc Thesis - Abstract
Characterisation of native tropical rainforest covering the study site, the Bobiri Butterfly Sanctuary, Ghana, has been most successfully completed in this exercise by assigning the HH, HV and HH-HV polarizations of the ALOS-Palsar instrument respectively to computer monitor RGB channels.
SRTM tile selection in Google Earth
This method required that ground sourced allometric data, which had been recorded with a certain projection system, was combined with the Palsar polarimetric data. This in turn required intermediate steps to prepare the data for the GIS software to combine all the data sets, using another image source for projection referencing purposes. A number of resampling transformations were used, with the Nearest Neighbour analysis appearing to offer the best solution in this case.
image pair composite
Imagery was compared visually with the Globcover land use classes, more precisely using side-by-side referencing by running Cartographica and Gamma X-window together, and finally by incorporating all the data sets together within a GIS, and experimenting with layer orders and transparencies on data with common projection systems.
transparent Nearest Neighbour over Globcover
From this, it is clear that a directly imported data source from Gamma Palsar processing software, into GIS RGB channels would be the simplest and most reliable, would it were possible to read-in the location information in the accompanying .par files.
RGB composite of nearest neighbour
Because of the importance attached to ground sourced allometric data in the formation of any forest type mapping exercise, the ALOS-Palsar instrument is employed because of it’s L-band forest cover suitability and the availability of site imagery.
CC Polarization RGB assigned composite over Globcover (overview)
Having contact with scientists working within the Bobiri Butterfly Sanctuary willing to forward this important source of ground data, enables the exercise of characterising the Bobiri site in a variety of ways, but a consistent methodology is required to enable a transferrable model that will produce a consistent output.