The Data Used

The project uses various sources of data to examine both the current distribution of land-use intensity in Vancouver as well as looking at relative potential to densify based on a set of criteria that match what was outlined as goals of EcoDensity and Smart Growth Theory.

The Data was derived from the following sources:

-        Census Data: DMTI Census Data 2001

o   Population

o   Total dwellings

o   Average number of bedrooms per dwelling

-        GVRD Spatial Data 2001:

o   City boundaries

o   Roads (major, minor, allroads)

o   Roads buffer polygons (major, minor)

o   Land-use layer 2001

o   Land-use layer 2006

o   Schools points (Secondary and Elementary)

o   Vancouver neighbourhood boundaries

-        Translink:

o   Transit stop locations

o   Transit lines

o   Average speeds of different transit types

-        Georeferenced:

o   Community centre locations (City of Vancouver website)

o   Canada Line and stations

 

---

Current Densities

The first step of the mapping was to examine the current density and land-use intensity patterns in Vancouver to create clear visual representation of the situation today.

There are in existence various maps showing basic density in Vancouver, and, whereas these are useful for a snapshot view of the city’s population distribution, absolute density measures fail to adjust for various variability within areas. In other words, whereas a pure density map would show high density in the West End and a low-density in Stanley Park, it would not take into account what those two land areas mean for possible population accommodation; that Stanley Park cannot accommodate any people, for example. The first step that this project intended to do was to look at the lands available for ‘population accommodation’ based on retaining our current land-use divisions (i.e. we are not going to build housing in a park). Land-use under any residential classification was therefore included along with ‘open and undeveloped’ since this land was assumed to be at least potentially convertible to future residential development. Only ‘Commercial-Residential/Mix’ category was somewhat difficult to include as it included large mixed areas surrounding the Central Business District (CBD) and especially along Broadway. These were partially excluded by using an older data set of Land-use from 2001which showed commercial specifically only along retail locations and not the entire block set as the 2006 did. The 2001 commercial polygons were erased from the ‘Commercial-Residential/Mix’ category of 2006 and the remaining areas were assumed to be generally residential areas. The 2001 land-use layer was not exclusively used at it had to many irregularities in other classifications especially in the Downtown area.

The following table shows the included and excluded land-uses:
 

Excluded Land-uses	Included Land-uses
Agricultural	Residential - Rural
Institutional	Residential - Single Family and Duplexes
Industrial	Residential - Townhouse and Low-rise Apartments
Industrial - Extractive	Residential - High-rise Apartments
Transportation, Communication and Utilities	Open and Undeveloped
Recreation and Protected Natural Areas	Commercial - Residential/Mixed*
Commercial
Harvesting and Research
Lakes and Water Bodies
Parks	*partially included

 

The next step was to remove the road areas with the assumption that these lands are not potential residential land. Existing road buffer polygon layers were used which had three different buffer polygon sizes depending on the road hierarchies (i.e. major roads had larger polygons). These road polygons were therefore removed from the residential land-uses using the ‘erase’ tool.

The remaining lands were considered the Residential Land Area, since this was all land that was potentially available for residential use in the city. A new field was created and the residential land polygon areas were calculated. The Residential Land Area layer was then joined with the Census DA layer for Vancouver using the ‘union’ tool. A table was created using the ‘Sum’ function which summarized the total polygon areas within each DA, and this was later ‘joined’ to the Census DA layer to finally give the total residential land area within each DA.

The first map that was generated was the Population per Residential Hectare (RHA) by DA which converted the area of residential land by DA into hectares and divided the DA population by that number. (See Map: Population per Residential Hectare)

The Second map was created using the same Residential Hectare data but looking at the number of dwellings registered on the census by DA per RHA. (See Map: Dwelling per Residential Hectare)

The third map used available Census data to calculate the number of bedrooms by DA by multiplying the Census data for ‘average number of bedrooms per dwelling’ by ‘total dwellings’ to determine the total bedrooms per DA. This was then divided by the population per DA to give us a map showing the average number of bedrooms per person by DA. Certain DA’s, however, were excluded since there seemed to be errors in the Census Data and these DA’s didn’t have any listed average bedroom value. (See Map: Bedrooms per Capita)

Current Amenities

The second point to examine was what infrastructure was in place to serve people in each area, with the assumption that increased density would benefit from certain existing infrastructure and amenities. Surplus population would therefore be better placed in areas which have more ‘amenities’ available. The amenities considered to be important were the following:

-        Parks

-        Schools (Elementary and Secondary)

-        Community Centres

-        Local Commercial Nodes

-        Public Transit

Schools were obtained from existing data in point shapefiles in the GVRD spatial data set. Community Centres were not available and therefore were manually georeferenced using the City of Vancouver website addresses and ‘Google Maps’ for location check. Parks were extracted from the Land-use layer and added as a separate layer.

Local Commercial Nodes were determined as commercial zone agglomerations throughout the City outside of the Downtown CBD area. Although for the initial Residential Land area selection, both land-use maps for 2001 and 2006 were combined to correct certain classification errors in the data, for the local commercial nodes only the older 2001 data was used because it provided separate residential and commercial classification in many areas where the newer map simply had ‘residential/commercial – mix’ zoned. Thus the commercial polygons for the land-use 2001 layer were selected and made into a new layer. Because the commercial zoning divided commercial zones into many tiny polygon areas and what was wanted was the commercial node areas outside the CBD, the polygons were thus aggregated using the ‘Aggregate Polygons’ tool under ‘Coverage Tools’. The aggregation parameter was tested over several distance settings to determine what would be appropriate in presenting an accurate picture of commercial node areas. The resulting aggregated polygons were checked with various known commercial areas to check for the most accurate distance parameter, which was thus set at 100 m. Also, the ‘orthogonal feature’ parameter was activated to maintain angular polygons and, finally, all polygons under 10,000 square metres were eliminated so as to only select the more major commercial nodes in the city with the assumption that these would perhaps be most appropriate for focusing densification around.

The final, and perhaps most important amenity examined for densification was Public Transit infrastructure. The data was obtained from Translink it and included a layer of all transit lines and stop locations. Each line was associated with a route number and these were classified into four transit types using the Translink website: ‘Skytrain’, ‘rapid-bus’, ‘diesel bus’ and ‘trolley bus’. Finally the future Canada Line Skytrain and stations were digitized using the locations based on the Canada Line website as this line will be completed soon and will greatly affect transit accessibility in that area.

Proximity Rasters

Once the amenities were selected and layers were created based on what was important in the criteria for potential densification, raster layers needed to be generated to look at relative proximity to these amenities with the intention of later running Multi-Criteria Evaluations.

The following Proximity Rasters were created:

-        Proximity to Parks

-        Proximity to Elementary Schools

-        Proximity to Secondary Schools

-        Proximity to Community Centres

-        Proximity to Local Commercial Nodes

Also, a complex cost distance path raster was created to look at:

-        Relative Time Proximity to CBD by  walking and transit

 

The initial step was to create a raster layer outline of Vancouver which was done by using a created Vancouver outline shape file, although this had to be modified to include areas where bridges crossed False Creek so as not to create a complete barrier here in our later transit analysis. All the subsequent rasters were limited to the extent of the ‘Vancouver_outline’ raster and were set at resolutions of 25 metres.

The first five proximity rasters (parks, elementary schools, secondary schools, community centres and local commercial nodes) were calculated using the ‘Straight Path Distance’ tool under ‘Spatial Analyst’. It wasn’t considered important enough to look at road geometry here since with walking and cutting through blocks and parks a more simple  straight-line proximity was deemed sufficiently accurate.

Transit Accessibility to the CBD

To calculate the transit accessibility to the Downtown, data on stop distances to Downtown was unavailable from Translink and therefore a somewhat complex model was needed to estimate times to the CBD for people walking and using transit - the forms of transportation which would want to be the focus of future densification rather than relying on auto-dependent development. The analysis performed was to use a Cost-Distance Path measurement to the Downtown using relative transit and walking speeds as the friction component (or cost).

 The initial layer that then needed to be created was a raster of the cost friction. The base of the raster was arbitrarily set at a value of 500 and that value was to represent the relative speed of walking.  Each of the transit lines were then given a relative cost compared to the 500 base for walking speeds. The relative costs were calculated using Translink data from a 2001 study showing average speeds of different transit types in Vancouver and including calculated stopping time. The transit types were ‘trolley buses’, ‘diesel buses’, ‘rapid-bus’ (such as the 99 and 98 B-lines) and ‘Skytrain’. The speeds were thus compared as a relative multiple of the average walking speed which was referenced at 5km/h. The following table shows the average speeds of each transportation type and the relative cost attributed:

Travel Form	Average Speed[1]	Relative Cost Friction Assigned
Walking	5 km/h	500
Trolley – Bus	13.5 km/h	164
Diesel Bus	15.2 km/h	116
Rapid Bus	21.6 km/h	100
Skytrain	40.3 km/h	62

 

The relative cost friction was calculated based on the walking speed and this was added in a new ‘Cost’ field in the transit lines layer based on each lines’ transit type. This transit line layer was then converted to raster with the Cost field as the new raster value. For the buses it was assumed that the transit stops were close enough to each other to assume that the overall line was the same friction value but for the Skytrains more steps were needed to ensure that the analysis wouldn’t assume that one could simply get on a Skytrain anywhere along its length and therefore an effective ‘block’ was needed along the line except at station locations.

The first step for the Skytrain was to digitize the Canada Line. As this line would soon be completed in 2009, it was important to look at its influence on transit accessibility to the Downtown and how the neighbourhood surrounding it might be potentially important for densification.

In creating the ‘block’, all three Skytrain Lines (Millennium, Expo and Canada Line) were buffered by 50 m to be large enough to be represented in the 25 m resolution rasters created. Another buffer of 20 m was then done to represent the actual line itself. This second buffer was then removed from the wider raster using the ‘erase tool’ and thus a polygon strip on either side of the Skytrain lines was obtained. The next step was to ‘cut’ the polygons in edit mode to cut out strips where the stations were located along the line and erase these; effectively leaving a polygon strip that follows the Skytrain along the line except where stations are situated. Converting these to raster and using the ‘Raster Calculator’ this Skytrain block area was then cut out of the Transit Cost raster layer.

The picture on the right shows the final Transit Cost layer with the blocks (the white gaps) along the Skytrain lines (shown in red) and each transit line is represented by an appropriate friction with the overall remaining land area at the friction of walking speed.
 
Once the final Transit Cost layer was accurately made, a ‘Cost Weighted Distance’ was calculated under the ‘Spatial Analyst’ tool set. A polygon of the CBD was used as the ‘target’ location. The resulting raster layer, therefore, showed a relative cost distance from any pixel on the map to the closest edge of the CBD polygon and provided a very useful visual tool of transit proximity to the Downtown core for all of Vancouver.

 

Multi-Criteria Evaluation (MCE)

Having created raster layers for all the current density maps and the selected amenities in Vancouver, the next step was to run the Multi-Criteria Evaluation and thus produce maps combining the information and showing the areas that best matched the criteria for potential densification.

The first step was to combine all of the beneficial amenities raster into a single MCE output which then could be used to compare to the current density levels. The first step in the MCE was to normalize all of the raster layers so as to convert the varying data to a simple 0-100 scale. This also involved reversing the values so that, for example, a low distance to a community centre would score a high number since it would be matching our MCE criteria the most closely. The simple formula used in the raster calculator to normalize all of the amenity rasters was the following:

 

(1 - [layer]/MAX) * 100

Where MAX represents the maximum value found in the layer

The weights were based on the assumption that the most important amenity factor was transit accessibility to the CBD as this was seen as the most difficult to change. It is feasible for example to add an extra school into a neighbourhood, however, much harder to make a distant area more transit accessible to the CBD, such as through building another Skytrain line. Along this same logic commercial nodes were seen to be the second most important as they are already developed secondary commercial centres outside the CBD and provide the perfect nodes to concentrate development. Parks are also considered very difficult to increase and existing parks provide a good green space amenity surrounding which density can be concentrated. Finally community centres, elementary and secondary schools were considered less central but still important amenities for increased population as it was generally assumed that, even their current capacity wasn’t able to meet increases in population in the area, it would be easier to expand capacity within existing school sites than building more schools on scarce land.

 

Amenity		Raster layer name	MCE weighted multiple
Transit accessibility		ntranscostd	10
Proximity to commercial nodes		distcomrnodes	5
Proximity to parks		distparks	2
Proximity to community centres		distcommcentr	1.5
Proximity to secondary schools		distschoolss	1.5
Proximity to elementary schools		distschoolse	1
	Total MCE Weights		21

 The following equation shows the full process of calculating the final MCE for amenities, including the normalization, the reversing of values and the relative weighting:

(1 - [ntranscostd] / 1715938.875) * (10 / 21) + (1 - [distcomrnodes] / 2926.175048828125) * (3 / 21) + (1 - [distparks] / 1131.923095703125) * (2 / 21) + (1 - [distcommcentr] / 4465.49267578125) * (1.5 / 21) + (1 - [distschoolss] / 3418.7900390625) * (1.5 / 21) + (1 - [distschoolse] / 2970.690185546875) * (1 / 21)