Methodology

Main Index

            I decided for my project that I would perform both a Boolean MCE (multi-criteria evaluation), and a weighted linear combination of my factors and constraints so that in the end I could compare the two resulting images to determine their differences and similarities.  First off, I began by preparing data for the Boolean MCE by using reclass, assign and buffering.

Preparation for the Boolean MCE:

Constraints

           First and foremost I decided to look at my landuse layer and break down the types of landuse into suitable (assigned with a 2) versus unsuitable (assigned with a 1) to produce the image LANDCON.  A value of 0 was assigned to the ocean because it is not even considered in this analysis.  The types of landuse I considered to be suitable were immature forest, barren lands, urban areas, estuaries, agricultural lands, and recreational areas because it is quite possible to live in any of these areas.  The areas deemed unsuitable included ocean, mature forest, logged areas, wetlands, lakes, unknown, alpine, and ice fields because they are inappropriate to live on under any circumstance.
 

The second major constraint in my analysis was related to distance from a populated city centre.  I consider this to be a constraint because living too close may be hazardous in an earthquake, yet living too far away will be costly and impractical.  To develop the final image CITY_DISTANCECON, I entered data on population densities as well as information on earthquake magnitude of previous earthquakes in those same cities into a datatable in database workshop.  These files were exported as .avl files and then assigned to the raster image major_cities which contained the locations of the seven major cities that I digitized in vector format and then rasterized.  A distance operation was performed on major_cities and then reclassed to produce the final image which states that the best place to live is at least 5 kilometres away from a major city, but not more than 100 kilometres away.

Factors

            I determined there to be five major factors affecting the safety of a location in an earthquake.  These are distance from lakes, distance from dams, distance from hydrolines, distance from roads, and slope.  Regarding lakes, dams and hydrolines, I surmised that the further away you live from these features during an earthquake, the better.  All three of these layers had to be rasterized using various conversion techniques (polyras, pointras, and lineras).  The raster layers were then reclassed and a buffer was put on each layer individually to reflect the necessary distance to be away from these features for safety.  For lakes I determined that a 2000 metre (2km) buffer would be appropriate, for dams, a 5000 metre buffer was used because they can cause a great deal more damage than a lake, and for hydrolines a 7000 metre buffer was used.  I figure that hydrolines are the most dangerous feature to be near in an earthquake, especially if you are in an area that is flooded, because you are sure to die from electrocution.  Once the buffered layers were produced they were each overlayed (using the AND overlay operation) with a reclassed raster image of Vancouver Island which served as a background on which to place the features.  Finally the resulting images were reclassed to produce lakebuffbool2, dambufferbool2, and hydrlinebuffbool2.

            To create allroadsbuffbool2, a few more steps were required.  First of all both the m_roads and isldhwy vector layers were rasterized using lineras, and then they were overlayed to produce an "OR" image which showed the similar features of both.  The m_roads image covered the isldhwy image to produce allroads which was then buffered to show areas within 1500 metres of any road.  This buffered image was overlayed with Vancouver Island, and then reclassed to produce the final boolean image.

            The final factor included in my MCE analysis was slope of the land.  To create this image I used the DEM image of Vancouver Island and performed a surface analysis on it in degrees to create a slope image.  This slope image was then reclassed so that areas with less than a 10 degree slope were considered suitable, and areas with a slope of 10 degrees or greater were considered unsuitable because there is an increased danger of a landslide occurring.  Again this reclassed image was overlayed with the Vancouver Island image with a multiplication "AND" overlay, to produce slopebool.
 

Preparation for the Weighted Linear Combination:

           In order to create the WLC, all of my factors needed to be converted into fuzzy images which have to be based on distance images in order to weight the individual factors at the end.  My two constraints need not be converted because they are boolean images and will be added as such in the WLC.

Creating Fuzzy Images

            For all three of the images, lakes, dams and hydrolines I again converted them from vector to raster and then performed a distance operation on them to create distance images.  The fuzzy images lakefuzz, damfuzz, and hydrlinefuzz were then created from these distance images.  I chose to use the Sigmoidal function for these three layers with the monotonically increasing alternative because for all of these factors, suitablilty increases with distance from the target.   The reason why I chose the Sigmoidal function was because all of these factors have very low suitability within a certain distance of the target which increases as you move outward until it reaches a point where the benefits of distance levels off.  For lakes I determined the first value to be 2000, just as it was previously with the Boolean MCE, and the second value to be 5000 because beyond 5000 metres from a lake the risk factor of damage due to flooding is fairly equal.  For example a house located 5000 metres from a lake will probably not receive any more significant flooding than a house 7000 metres from the lake.  For dams the values I chose were 5000 and 10,000, and for hydrolines I chose the values 7000, and 12,000.  In all three cases, the first values correspond exactly to the values used for buffering in my prior analysis.

            To create slopefuzz I performed a distance analysis on my slopebool image described previously to create slopedist.  I then created a fuzzy image again using the Sigmoidal function, yet this time it was monotonically decreasing.  I decided that areas with a slope less than 10 degrees are suitable so as you move away from these slopes, suitability will decrease.  My "a" and "b" values were then 0 and 10 respectively.

            Creating allroadsfuzz was a little different than with the other images because this time I used a monotonically decreasing  J-shaped function because the most suitable areas are near to roads.  With the J-shaped function, suitability away from roads decreases continuously, but never reaches a value of 0.  My first control point was 500 metres (most suitable area), and my second control point was 1500 metres (the value at which suitability is halfway between not suitable and perfectly suitable).

Final Images:

            The final step was to create both the final boolean MCE image (finalbool) and the final WLC image (solution)and compare the two.  These final images can be viewed in the spatial analysis section.

Main Index