Summary of the intensification analysis
The following is a summary of two technical papers that are currently in progress. The first will document methods used to derive a 1990 built-up urban boundary from satellite imagery, and the second will describe the use of census data to estimate average intensification rates from 1991 to 2001. These papers will be published and distributed through a working research paper series website hosted by the Geography Department at the University of Toronto. The methods described below and in the technical papers have been presented to and commented upon by academics in the fields of image processing, geographic information science, and urban geography.
The objective of the research was to develop a method of estimating recent historical rates of intensification that would mimic the approach proposed in the Growth Plan for the Greater Golden Horseshoe and its supporting documents, developed by the Province of Ontario. Building on research undertaken in 2004, Neptis derived a 1990 boundary for the built-up urban area in the Greater Golden Horseshoe (GGH). The analysis was to accurately and conservatively identify the consolidated urban area in the region, as opposed to small, scattered urban features located far from large contiguous urban areas. Imagery captured by Landsat Thematic Mapper 5 (30-metre resolution) in summer 1990 was selected for the urban boundary analysis. Several image enhancement techniques were performed, such as texture analysis, normalized difference vegetation indexing, principal component analysis, and image differencing, in order to employ more than the spectral values of the imagery in the analysis. A supervised classification was conducted using a Bayesian Probability Function calculated from the inputs for classes established from training sites. Each pixel is assigned to the class to which it most probably belongs. The classification produced a data set with three categories: water, non-urban, and urban. The overall accuracy of the classification was 98.33%, with a 91.04% user's accuracy for the urban class, meaning that over 91% of the pixels classified as urban correspond with urban features on the ground.
The 1990 urban area boundary was then used with 2001 census geographic units and two census variables to estimate intensification rates for all upper- and lower-tier municipalities in the GGH. Initially the 1990 urban boundary was used to select all dissemination areas (DAs) that intersect the boundary. The analysis captured not only DAs that are completely contained within the 1990 urban area, but also DAs that straddle the boundary. For these DAs, the period of construction census variable was used to sum all dwellings built between 1991 and 2001. To identify the number of dwellings that fall outside the urban boundary, census blocks (a very small geographic unit whose boundaries correspond to neighbourhood blocks) were employed. The 2001 census enumerated all dwellings for each census block. A second intersection analysis was used to identify all blocks of which more than 50% fell outside the 1990 urban area. The dwellings units were summed for these blocks. The block sum was then subtracted from the DA sum to yield all dwellings within the 1990 urban area that were built between 1991 and 2001.
Next, a buffer analysis was performed, which calculated a 500-metre distance inside the urban boundary. DAs of which 90% or more was within the 500-metre buffer were identified. The period of construction variable was summed for these selected DAs, yielding the number of dwellings built between 1991 and 2001 within 500 metres of the urban boundary.