Problems and Discussion
Some of the main issues that were encountered in this study involved missunderstanding of the software. First, I presumed that when importing vector files into IDRISI the software recognised the incoming reference system and that chosing a reference system at the import stage was a form of conversion. This was not the case. The evidence of this error did not manifest until the final map was produced and the two data sets did not line up. In attempting to rectify the problem (redoing all the analysis) a larger issue came to light. IDRISI has a canned set of projections from which to choose from when importing. The data for this project were in a specialized projection for B.C., Alberts Equal Area Conic. This particular projection was not in the selection available in IDRISI. Upon attempting all the Alberts projections that IDRISI offered I realized, of course, they did not work. Clearly projection is an absolutly vital issue in GIS. Further education on the matter would be very valuable.
Secondly, I originally did a non-boolean weighted linear combination for the final stage of the analysis. I presumed that this functioned like an "and" question, that it weighted areas that contained both criteria. In fact the results returned a map that depicted non-forested areas close to the sea as habitable. Thus this function trades in an "or" fashion where one criteria can entierly superseed another criteria. This dilema could be remedied by doing a boolean overlay after the fact eliminating non-forest. However I felt I would get a better result if I did not perform a weighted linear combination and rather did a simpler boolean operation "First * Second". In saying better result I mean a result which I have a better comprehention of.
Another issue in data preperation was how to go about seperating information from tables so that it could be classed. In IDRISI, atribute values files can be developed from a fields table and by this data can be reassigned. I was not aware of this at first so I queried the data out of tables in ARC/VIEW and created seperate shape files of the forest types, imported it to IDRISI and then converted to raster. This created a situation where the BEC zones were no longer mutually exclusive. The rasterisation created jagged edges that overlapped slightly when the shape files were overlayed. This problem made boolean functions useless because overlays returned values of more than one.
This study was severly limited by data storage constraints. I had to cut the study area down to the area around Vancouver Island only because the data set far exceeded the space which was aloted in the lab. The actual habitat of the Marbled Murrelet ranges almost the entire Pacific Coast.
A final word, although this project has felt like some form of torture or a hazing ritual in that it has been very demanding, it has been an invaluable learning experience. In hind sight, the actual analysis now seems rather simple it was getting to this point that was the challenge.
