The in depth spatial analysis of this project, is focused on the municipality of Burnaby.  By assessing what type of landuse the area had, along with the existing malls, and the road network, a more accurate location for a new Gap can be found.  Furthermore, the above criteria can be better assessed by using fuzzy logic to add friction into a cost analysis, and weights to a MCE.   By using range defined constraints, a more accurate suitable area will be defined than if a boolean constrain were run, as boolean constraints only view the criteria as being acceptable or unacceptable.  With range defined constraints, however, there is a range of suitable outcomes.  Moreover, the range can be weighted and ordered depending on which criteria are more important to the question at hand.  Click this link to view a comparison outcome between the two aggregation methods.
    Before I could begin any spatial analysis, though, I had to create a secondary data source from the GVRD raster image.  I was introduced to the operation called window, which allowed me to essentially "cut" Burnaby from the original image.  However, before I did that I performed an overlay, using linras to rasterize the vector image called major roads, as I knew that I would require this information later.  So, the procedure was simply to reclass the GVRD image, in order to accept only the municipality of Burnaby as the boolean criteria.  This left an image of mainly black area covered by roads, with only a small Burnaby area, covered by roads of a different colour.  From this image, it was easy to boolean out the extra roads, outside of the Burnaby area, as they were already classed uniquely. With the extra roads gone, I initiated the window operator, and set the parameters to the area encompassing Burnaby.  The resulting image offered a close up view of Burnaby with all of its major roads.

    Next, I needed to add the existing mall information, for Burnaby.  I began to digitize the malls as points on a new vector layer (over top of my Burnaby-Roads image, for reference), by combining the address information from the CSCA excel file with a BC Road map.  Then the new mall vector file was rasterized via pointras, and finally overlayed with the Burnaby-Roads image to create the project location area. Now, all that was left was to gather the Burnaby landuse information, which was already provided via the Goeg.355/data files.

    So, now that I had created the roads, malls, landuse, and Burnaby area files, I employed fuzzy logic starting with landuse.  Commercial sectors were defined as being highly suitable areas, with open undeveloped space considered middle ground, and industrial sectors considered low, but still feasible.  Once I had created the above, called LanduseDist, I still needed to use the landuse file to act as the friction base with the mall file.  When friction was run with malls, a cost distance surface was generated and then the operation Fuzzy was ran on it, using the highest and lowest values to cross the cost distance image.  These analysis produced a final file, called costdistfuzz, that was also on a range scale where 255 equaled the closest suitable landuse areas from the malls, with the areas going down in range to zero equal the more "expensive" areas, or unsuitable landuse areas outward from the malls.  The fuzzy logic procedure was run on malls to generate a equivalent range scale where 255 equals the mall site, and as the scale decreases, so does the distance from the existing mall.

    The above three transitions into a range scale relied on linear fuzzy logic, as the value of the distance from the source decreased uniformly.  However, the distance from accessible roads within the municipality followed a j-shaped logic, in that as distance increased from the source, the value decreased, but never reached zero.  To clarify, this logic assumes that people would travel outside of this municipality if they were sufficiently motivated.  As a result, when the fuzzy operator was run on roads, j-shaped logic was used with the first control being 50m from the source and the second constraint being 600m from the source, producing roadfuzz.

    Finally, to put all of the information together, I ran MCE (Multiple Criteria Evaluation) using both a Weighted Linear Combination, using the following weights:  landuse - 0.5346, mallfuzz - 0.2716, costdistfuzz - 0.1460, roadfuzz - 0.0477.  As well, I ran an Ordered Weighted Average using the following order:  mallfuzz - 1, costdistfuzz - 2, landuse - 3, and roadfuzz - 4, because I felt that the most important criteria to placing The Gap was the distance from the existing malls, and the distance to the surrounding suitable landuse.  The end result, no matter which order the criteria was in, yielded the same results.  Now, that all of the information had been analyzed, it was time to draw some conclusions from them.  Click the link below to read the conclusions.