Methods

The plan of the project was to create a health index map for the Greater Vancouver Regional District. Using the research ideas from the many articles related to healthy environments I decided a Multi-Criteria Evaluation would be the best way to create a health index map. The key steps were to create normalized indexes of each of the factors. Once the Euclidean Distance maps were created I then applied a Multi-Criteria Evaluation, the weighting factor was determined through the many studies done in this field. The final map would show the level of health in the environment, and with the census data one would be able to overlay the two and hopefully discover patterns.

Projections/Extents

The projection used for this project was NAD 1983 UTM Zone 10N. The extent for the raster maps created was limited to the GVRD. The GVRD boundary was established through the use of the layer (cities) from the G: drive in the GVRD geodatabase.

Data Collection/Manipulation

A list of all the data collected can be found here. All the data was ‘clipped’ to fit only in the GVRD boundaries. The layer (boundaries) was used from the GVRD geodatabase to clip the data. A lot of the data was merged to create features that included only the necessary components. The merged data includes:

The fast food and grocery store locations were collected through the use of Google Maps. The address and city locations were collected and geocoded into the final map.

Euclidean Distance Maps

Euclidean Distances for the point and polygon feature classes of Community Facilities, Fast Food Restaurants, Grocery Stores, Leisure Activites, Parks, Recreational Centres, and Transit Stops were created. The ‘Euclidean Distance’ tool simply creates a raster map showing the distances deviating from the point, line, or polygon source. Due to the extent set, the raster maps would be limited to the GVRD. I used a 5 class scheme for each raster map marked at the 400, 800, 1200, 1600, and greater than 2000 meter markers. Here is the link for a raster map created for transit stops.
The choices for the distances were based on the many research articles. According to some of the research, a person is willing to walk to transit stops, parks, and other amenities as long as it is within roughly half a mile (≈800 meters). By setting up the classification scheme in this form, one can get a sense of both distances people are willing to walk and distances where they would opt to drive. Since the application is to the GVRD I changed the maximum limit to about 1600 meters. Within my own commuting experience, I know I walk about 1600 meters to get to the closest transit stop and I also believe most people in the GVRD are also willing to walk to transit stops, parks, and other amenities.
Of course this is an assumption and could possibly be a source of error, especially considering the variability of the people. Some people may enjoy walking or some may consider the walk to the park a part of their workout regime. Considering all these sources of deviation I believe a maximum walking distance of 1600 meters is a fair assessment.
Once the raster maps were created, I extracted the raster values by using the mask layer (cities) to only get values for the GVRD and not the surrounding areas, such as bodies of water

Index Maps

All the raster maps created through the ‘Euclidean Distance’ tool were then converted to a normalized index. Basically all the values of the raster were converted to values between [0, 1]. The value of 0 represented areas beyond the maximum walking limit (1600m), while values of 1 signified areas that were right beside the source feature.
The ‘Fuzzy Membership’ tool was used with a linear transformation. The minimum value was 1600m and the maximum value was 0m. Here is the link for the index map created for transit stops.

Grocery Store/Fast Food Analysis

Similarly with the grocery store and fast food location data, Euclidean distance raster maps were created. Again with the exact classification scheme as in the other source features, and this map was also extracted by the mask layer (cities).
Using the ‘Raster Calculator’ tool, I created the condition equation: “Con(GroStoDist > FastFoodDist, 1, 0). Basically this equation says that if the grocery store distance is greater than that of the fast food distance; assign those areas with a value of 1 and 0 for the others. The values of 1 show areas where access to fast food restaurants is easier and values of 0 show areas that are closer to grocery stores. This map will not be included in the MCE, the sole purpose of this map is to give a visual of which areas are closer to fast food restaurants and vice versa.
Finally, a normalized index was created for the grocery store distance data using the ‘Fuzzy Membership’ tool. Again the maximum distance limit used was 1600m, and this index will be used in the final MCE. Since the goal of the project is to create a health index, between grocery stores and fast food restaurants only grocery stores will be used because it is the healthier choice.

Multi-Criteria Evaluation

The tool used to process the multi-criteria evaluation is called ‘Weighted Sum’. Basically this tool takes in the normalized factors with their respective weighting and creates a raster map. The map I will create will show the final health index of the region. Since the results of a multi-criteria evaluation are directly dependent on the weighting given to the respective factors, I will be conducting a sensitivity analysis. Essentially, I will be applying multiple weighting schemes and looking at discrepancy of the results and how the model is affected by changes and assumptions.

The first multi-criteria evaluation result will be based on the perspective of a public transit user. For this reason, the accessibility to transit stops will have high weighting. People dependent on public transit need easy access to transit stops or they will not be able to commute anywhere. This could mean that these people are unwilling to walk to parks, recreational centres, grocery stores, etc. They may just resort to what is close; if their environment is full of fast food restaurants this could have a detrimental effect on their health. Here is a table showing the factors and their respective weighting:

Factor Weighting
Community 10
Grocery Stores 36
Leisure 3
Parks 14
Recreation 7
Transit 30



The second multi-criteria evaluation result will be based on the perspective of a non-transit user. For somebody who owns a car it is unlikely they will go shopping or commute anywhere without their car. For this reason, a non-transit user will have less weighting on accessibility to transit stops, as well as proximity to grocery stores. The left-over weighting would be put into areas such as accessibility to parks, recreation centres, and leisure activities. Here is a table showing the factors and their respective weighting:

Factor Weighting
Community 20
Grocery Stores 3
Leisure 8
Parks 40
Recreation 25
Transit 4



The last multi-criteria evaluation result will just be an equal weighting result, each factor having a weight of 16.67. This will then be compared to the other MCE’s to see how much the end results differ. This MCE is for the sensitivity analysis.

Census Data Analysis

After creating the MCE maps, I used 2006 census data to check who lived where. The data was obtained from the UBC Department of Geography. Dissemination areas (DA’s) were used to aggregate the GVRD, and the DA map for the GVRD was also provided by UBC.
The first step was to join the age/sex and household income tables to the GVRD DA map. Once the tables were joined to the DA Map, I exported two different maps. One map had age/sex data and the other had household income data.
Then using the ‘Selecting by Attributes’ tool, I was able to find the DA’s where at least 20% of the population were children aged 0-14. Using the same method I was also able to find DA’s where at least 20% of the population were elderly aged 75 and beyond.
With the household income data, I found the DA’s where the median household income was less than or equal to $25000/year before taxes.
Using these new DA’s I was able to overlay the DA map with 50% transparency over the MCE maps and see where these people were living in accordance to the level of health of their environment.