Click here to see the cartographic
model of this project
.
The first step in analyzing this problem it to calculate the distance
from each pixel to each of the 9 criteria for evaluation. To calculate
this, the DISTANCE function was run on each layer. Here is a sample
of the results:
Distance from Bike routes:
Distance from a firehall:
Distance from Shopping area:
For the first kind of MCE, I decided to find all of the areas that contained
at least 5 of the 9 criteria. To do this, I converted each distance
coverage to a boolean image, based on my (somewhat arbitrary) conditions,
and then added the images together using the OVERLAY function. The
conditions that I used are as follows:
- The site must be between 100 and 1000 meters from a main road
- The site must be less than 20m from a side road
- The site must be within 100m of a bike route
- The site must be between 100 and 500 meters from the Skytrain
route
- The site must 50 and 700 meters from a park
- The site must be within 200 meters of a shopping district
- The site must be between 50 and 1500m from a fire department
- The site must be between 50 and 500 meters from a bus route
- The site must be between 50 and 500m from the water
Here are some results of applying these conditions to the distance coverages.
They were obtained by using RECLASS to assign the desired values 1,
and the undesirable values 0.
Suitable distance from water:
Suitable distance from a bus route:
Suitable distance from a side road:
I then used the following macro to perfrom the OVERLAY addition to add
the layers together:
overlay x 1*mainroadbool*sideroadbool*temp1
overlay x 1*temp1*bikebool*temp2
overlay x 1*temp2*skytrnbool*temp3
overlay x 1*temp3*parkbool*temp4
overlay x 1*temp4*shopbool*temp5
overlay x 1*temp5*firebool*temp6
overlay x 1*temp6*busbool*temp7
overlay x 1*temp7*waterbool*overlay-final
The result of the operation is below. I then did another RECLASS
operation to reassign a new value of 1 to all squares where the boolean
add resulted in a value of 5 or more.
Original overlay:
Reclassed Overlay to show most suitable areas:
For the next type of MCE I performed, I needed to reclassify the distance
images using the FUZZY command. The specific functions that I used
are outlined below. The main difference between J shaped functions
and Sigmoidal functions is that Sigmoidal functions reach 0 at one of the
control points, while J-shaped functions are asymtotic to 0 (they never actually
reach 0), and the control points represent the point where suitability is
0.5 on a 0 - 1 scale.
- Mainfuzz - decreasing Sigmoidal - control points (c, d) at 100m,
1000m
- Sidefuzz - decreasing J-shaped - control points at 20m, 40m
- Bikefuzz - decreasing J-shaped - control points at 100m, 300m
- Skytrnfuzz - Symmetric Sigmoidal - control points at 0m ,250m,
250m, 500m
- parkfuzz - Symmetric J-shaped - control points at 35m, 50m, 500m,
550m
- shopfuzz - J-Shaped - control points at 200m, 300m
- firefuzz - decreasing Sigmoidal - control points at 50m, 1500m
- busfuzz - J-shaped - control points at 35m, 50m, 300m, 350m
- waterfuzz - Symmetric Sigmoidal - control points at 35m, 50m, 500m,
1000m
All the functions were rescaled on the byte scale, to give a value of 0
- 255. Here are some selected results:
FUZZY applied to the Main road distance coverage:
FUZZY applied to the park distance coverage:
FUZZY applied to the Skytrain distance coverage:
After all of the FUZZY images were created, used the WEIGHT function to
determine the wieghts on the various criteria. I generated 2 different
weight files, based on 2 different types of theoretical people, who are looking
to move into the city. The first person is an outdoorsy-type, and
had preferences as follows:
- Must live near a bike route, since she commutes by bike to work
- Likes to walk in the park, so distance to that is also important
- Doesn't own a car, so close to transit is also important
- Would prefer to live away from a main road and shopping areas,
to cut down on noise/crowds/pollution
- Wants to be a reasonable distance from a firehall, for safety's
sake
Based on these preferences, the weights turned out like this:
The eigenvector of weights is :
mainfuzz : 0.0471
sidefuzz : 0.0695
bikefuzz : 0.2393
skytrnfuzz : 0.0965
parkfuzz : 0.1624
shopfuzz : 0.0436
firefuzz : 0.0962
busfuzz : 0.1505
waterfuzz : 0.0949
Consistency ratio = 0.09
Consistency is acceptable.
Running MCE with these weights produces the following output file, which
I think is neat-looking:
For the second evaluation, I adjusted the weights to more suit the following
hypothetical person:
- Is a shopaholic, so proximity to a main shopping center is paramount
- Needs to live near a main route to get to work/mall
- Would like to be close to a firehall, for safety's sake, just
like the other person
- Doesn't use transit or ride a bike, so those things are not important
Based on these preferences, the following weights resulted:
The eigenvector of weights is :
mainfuzz : 0.1886
sidefuzz : 0.1087
bikefuzz : 0.0442
skytrnfuzz : 0.0500
parkfuzz : 0.0630
shopfuzz : 0.2278
firefuzz : 0.1762
busfuzz : 0.0620
waterfuzz : 0.0795
Consistency ratio = 0.05
Consistency is acceptable.
Running MCE with these weights produced the equally compelling image below:
My final step was to RECLASS both of these images to identify spots where
the suitibility was above a certain value. I choose 175 out of 256
as this value and reclassified the images as such. Here are the results:
RECLASS of the Outdoor person's map:
RECLASS of the
Indoor person's map: