Part 1: MODIS Raw to Final Processing
Initially, large range data processing was attempted, starting with MODIS white and black sky spectral albedo and ending up the broadband blue-sky albedo. The processing consisted of several steps.
Step one: Obtain data
Data was ordered through NASA’s Land Processes Distributed Active Archive Center (LP DAAC). Initially the black and white sky albedo parameters over a 16-day period at 500 meter resolution combined from the terra and aqua satellites (Data product MCD43A3) was selected for the south east portion of the United States. The data was transferred via browser download (as opposed to ftp pull). The approach to processing the data consisted of importing the data into some useable form, performing mathematical matrix operations to process the data to final form, and exporting the data into a final geo-referenced dataset. Several software applications were considered to accomplish this, including Octave, R, MATLAB, ArcGIS, and Erdas Imagine. After some initial investigation, it was decided that MATLAB would be used for the data processing and ArcGIS would be used for the final geo-referenced display.
Step Two: Initial Importation
In order to get an idea what the data was like the raw data was first imported into ArcGIS. Land product data sets from MODIS are encoded in HDF V005 format, which is a self describing format that is said to have wide support. Unfortunately, HDF5 is not well supported for importation with ArcGIS. To get around this dilemma, MODIS products were first imported into a remote imaging program, Erdas IMAGINE, converted to an .img file, and then accessed directly in ArcGIS. While the program was excellent with dealing with the GIS side of the project, it was decided that it would not be the best program for the actual processing because its matrix manipulations and scripting are based on Python (which I am not familiar with).
Step Three: MATLAB processing
Fortunately, HDF5 is supported with MATLAB’s universal importation tool, and the matrices were introduced into the MATLAB workspace with no problems. A MATLAB m-file (BS_Albedo.m) was written that uses the data, along with input fraction of diffuse skylight and optical depth. Since this was based on the white and black sky albedo data products (which assume solar zenith angle to be at local solar noon) the data was only considered a rough estimate of actual blue sky albedo; the m-file was not tested and only considered to be in an alpha stage of development. Another m-file was written to take these spectral values and convert them to visible broadband albedo with published weight parameters. Again, this m-file is only in alpha stage of development and not tested.
After creating somewhat dubious results, the ORNL DAAC site was found, which can handle the raw to final data processing. Subsequently this approach to the project was abandoned.