Appendix A: Methods

Overview

First, we determined the urbanized area of each region in 1991, 2001, and 2011 from satellite imagery to calculate the amount and rate of urban area growth over 20 years. To measure growth in urban land, we delineated the built-up urban area in 1991 through the analysis of satellite imagery, which was subjected to a thorough visual inspection. We identified the expansion of this urban area between 1991 and 2001 using visual inspection of 2001 imagery and identified urban expansion from 2001 to 2011 using visual inspection of 2011 imagery.

This is a slightly different method from the one that was used in Taylor and Burchfield (2010), where the entire urban footprint was delineated for 1991 and for 2001 using a semi-automated method. The difference (subtraction) in the two urban footprints was used to calculate urban land growth between 2001 and 2011. The urban land increase (1991–2001) calculation in Taylor and Burchfield (2010) included new urban areas developed within the 1991 built-up area, or urban infill. In the current paper, our analysis of urban land growth focuses on the expansion area to calculate more precisely how much urban land growth is attributed to greenfield development.

Our subregional analysis focused on the decade of growth between 2001 and 2011. We used the 2001 urban area (the 1991 urban area plus the urban expansion area for 2001) to measure growth through intensification and the urban expansion area delineated from the 2011 imagery to measure growth through greenfield development.

Next, we overlaid 2001 and 2011 census dissemination area geography containing data on population and dwellings over the urbanized areas for the two years. By subtracting the number of people and dwellings in 2001 from the 2011 figures, we determined the net changes in population and dwellings over the decade.

We calculated average regional and subregional densities by dividing population or dwellings by the extent of the urbanized areas, and average regional and subregional household size by dividing population by the number of dwellings in an area.

In addition, we identified planning policy areas, such as major transit stations, urban growth centres, and frequent transit networks, to determine net changes in dwellings and population in these locations. We estimated the population and dwellings in these areas using census dissemination areas that intersected (overlapped) with the policy areas. In most cases, the census geography extended beyond the policy area boundaries, so in most cases the population and dwelling counts are an overestimate.

The spatial analysis methods used for the current report are similar to the spatial analysis methods used for Growing Cities by Taylor and Burchfield (2010), although there are some important differences. The original method (Du et al. 2007) used long-form census data from 2001, which includes a variable of the year of construction for residential units. This variable allowed the authors to identify units built in the decade between 1991 and 2001. However, the long-form census for 2011 has been replaced by the National Household Survey, which has been shown to yield less-than-accurate results for variables that are not part of the 2011 short-form census.

Therefore, we decided to limit ourselves to the short-form census. While the short-form census is a complete picture of population and dwelling type, it does not include the date of construction for residential units. Instead, we subtracted the total number of people and dwelling units present in 2001 census dissemination area geography that coincided with the 2001 urban area from the census dissemination area geography from 2011 to arrive at a net change in population and dwellings for intensification. We performed the same analysis for the census dissemination geography that coincided with the 2011 urban expansion area. In some ways, this approach should be more accurate than the earlier method, because the short-form census attempts to survey the entire population, rather than a 20% sample. However, the subtraction method is more complex spatially, as we had to align two sets of census boundaries, some of which had changed over the decade in question.

By using the subtraction method, we can see both population loss and gains in different areas. A detailed description follows.

STEP 1: Map urban expansion

We used satellite imagery to map areas that were urbanized in 2001 and 2011. Urbanized areas are identified by the presence of hard surfaces (such as buildings or pavement), as found in residential, industrial, and commercial areas. They do not include large green spaces such as parks, golf courses, or rural areas containing farms.

Although we started with the 1991 and 2001 urban footprints, which had been mapped for the original Growing Cities report, we modified them to match our interest in capturing urban edge expansion so that we could identify greenfield development by removing rural hamlets from the data sets.

Satellite imagery was obtained from the United States Geological Survey (USGS), which takes regular satellite images of the earth at a resolution that allows us to differentiate urbanized areas from nonurban areas through colour spectral bands. Newly urbanized areas were identified visually. Where necessary, we confirmed urban classification through examining higher-resolution imagery from Google Earth.

In the original Growing Cities report, both footprints were created using a semi-automated process, which allows for a consistent and repeatable approach, but may include errors of classification. For this reason, we relied on manual visual identification and inspection using the satellite imagery in a GIS program. This approach allowed us to avoid the misclassification of areas that appear urbanized, but are actually not developed, for example, land cleared for construction, gravel pits, roads in rural areas, and trailer parks.

Once the 1991 and 2001 urban area data sets were finalized, we used the visual inspection method to identify urban expansion between 2001 and 2011.

STEP 2: Overlay census data

Once we had determined how the urban footprint had grown between 2001 and 2011, we used census data to track population and dwelling changes over this decade.

Using the smallest census areas available, called dissemination areas (DAs), we started by overlaying data from the 2001 Census. Then we added a layer of DAs from the 2011 Census and compared the boundaries of the two. In some cases, DAs had been added by splitting old DAs; this is common in areas of rapid growth, especially in urban edge areas where new dwellings have been constructed.

DA boundaries are chosen to capture approximately similar numbers of people in each area; therefore, large DAs typically contain lower-density development such as rural or industrial areas, while small DAs are found in high-density urban areas. Most DA boundaries in existing urban areas remained unchanged over the decade.

STEP 3: Reconcile geometries in the expansion area

The third step involves reconciling the census geography with the urbanized area geography. Urbanized area polygons have different shapes and boundaries from those of the dissemination areas (DAs), which in turn can change when their boundaries are redrawn between censuses. In this study, the area with the most complex geometric differences was the expansion area around the edge of the existing urban area.

We overlaid the 2001 and 2011 DAs on the 2011 urban expansion area to determine which dissemination areas overlapped with the 2011 urban expansion area. We selected all overlapping DAs. These two overlapping layers of census data would provide us with before-and-after numbers for population and dwelling units for the decade 2001 to 2011.

Subtracting the total population and dwellings in 2001 from the total in 2011 gave us the net difference in population and dwellings in the expansion area. In order for the calculation to be accurate, the area covered by the DA need to be as closely matched as possible, given changes in DAs over time.

To reconcile these differences, we used an iterative process of adding and removing DAs from the selection to arrive at closely matched layers. In some cases we had to include parts of rural areas adjacent to the expansion areas that had a few dwellings or population. Doing so helped us to match the geometries more closely. These selections were checked and double-checked repeatedly, by different people, to ensure the most accurate match possible.

In other cases, we needed to include areas along the edge of the 2001 urban area. Because we were subtracting existing 2001 populations and dwellings to arrive at net changes in population and dwellings, including these areas did not affect net population and dwelling counts for the expansion area, as we can assume that most new dwellings in the suburban edge areas would have occurred in greenfields. Previous Neptis studies have identified delayed greenfield development at the edge of the urban area whereby small pockets of undeveloped land at the edge were more accurately characterized as opportunities for greenfield development than intensification (Burchfield 2010).

The process of matching the two layers of DAs to represent the urban expansion area was complete when we had as close a match as possible and discrepancies were minimized.

STEP 4: Identify areas within walking distance of frequent transit

We were also interested to find out how much growth was within walking distance of transit lines with frequent service. We defined walking distance as 500 metres to local bus or streetcar lines, or a 1-km radius of GO and subway stations. We included only routes that run frequently, every 15 minutes or less from 7 a.m. to 7 p.m. on weekdays. (GO train service does not yet fall within this frequency of service, but the Ontario government has plans to move to more frequent service.)

We identified census DAs the centre of which was within this walking distance, and labelled those DAs as being close to frequent transit (see Appendix 3 for DA selection for the FTN for each city-region).

We matched the geographies of the DAs over time using the same iterative process described in Step 3.

STEP 5: Identify areas in designated Urban Growth Centres

The Growth Plan for the Greater Golden Horseshoe identifies Urban Growth Centres (UGCs) as the focus of special intensification efforts in urban areas; Vancouver also has designated urban centres to focus development. We therefore identified which dissemination areas (DAs) overlapped with these centres. This step allowed us to calculate how much growth had taken place in these areas (see Appendix 3 for DA selection for UGCs for each city-region).

Some Urban Growth Centres had very complex boundaries, so we decided to include all the DAs that overlapped with UGC boundaries rather than just the ones with the centre inside the UGC as we did for the frequent transit network area.

The geographies of the DAs over time were matched using the iterative process described in Step 3. In Toronto, this sometimes meant that areas near the UGCs but not in them were included. In Vancouver, the urban centres’ boundaries were close to DA boundaries, but the same iterative process was required to ensure consistency.

STEP 6: Identify areas around GO stations

Using an 800-metre radius, which is how Metrolinx defines its station areas, we selected the DAs the centres of which were within this area. In some cases, we extended the area to ensure the DA boundaries matched over time.

Large DAs are usually non-residential areas with few dwellings. DA boundaries are delineated by Statistics Canada to ensure a comparable number of people inside each boundary. Therefore large DAs in and around GO stations do not necessarily include many extra people.

The geographies of the dissemination areas over time were matched using the iterative process described in Step 3.

STEP 7: Extract the data and calculate net changes in population and dwellings over the decade in different areas

Once we had categorized each dissemination areas (DA) as being either inside the 2001 urban area or in the 2011 urban expansion, near transit or not, and overlapping an Urban Growth Centre or not, we imported the data into Excel spreadsheets and organized them using pivot tables.

The data are broken down by municipality, by location (in the expansion area or in an intensification area), and by proximity to transit or Urban Growth Centres. We can also break down the data by population, dwelling count, and dwelling type in four categories.

Using the numbers from the two censuses, we calculated the net changes in population and dwellings over time in different areas. To do this, we subtracted the total population and dwellings in 2001 from those in 2011 to arrive at the net difference.

This calculation showed how much growth occurred in the existing urban area compared with the expanded urban area, how much had occurred within walking distance of frequent and rapid transit, and how much in areas designated as Urban Growth Centres. We can also determine how much growth occurred in the metropolitan region as a whole, and within each municipality.

Additionally, the surface areas of the 2001 and 2011 urban footprints were extracted from the GIS software and organized using pivot tables. This step allowed us to calculate population and dwelling densities for subregional areas and the region as a whole.

All numbers in the report have been rounded to the nearest ten to avoid a false appearance of precision.

The spatial analysis methods have been peer-reviewed by academics and a representative from Statistics Canada.