Apply the Model to the City of Montreal to Identify Census Tracts with the Least Tree-Cover and Greatest Heat Vulnerability
Objective three is to apply the model to the study area. Data on age, income, living alone, ethnicity, education, and housing structure are available at the census tract level. One of the key variables in determining areas of heat vulnerability is the proximity to cooling centres within the City of Montreal. The geographic location of recreational centres, public pools, libraries, splash-pads, and other facilities that can limit heat exposure to the public undergo a Euclidean distance analysis. The resulting proximity to cooling centres raster undergoes zonal statistics, using census tracts as the zone input. The output raster is aggregated to the census tract polygons, and the average distance to the nearest cooling center is assigned to each tract.
The tree-cover layer is derived from LiDAR data. The building and ground points were classified and removed using preset classification tools. The remaining points were subject to a height filter which removed points that were lower than 1.5 metres from the ground. What remained were point cloud clusters representing high vegetation. Using a point cloud to raster function, a 1 metre DEM derived solely the remaining tree point cloud was output. Due to the large nature of the LAS data, several iterations are run and the digital elevation model outputs are mosaicked. Once mosaicked, the DEM is subject to a binary classification of 1 for data present or 0 for no data which outputs a tree footprint raster for the entire city. A low pass neighbourhood filter is applied to smooth the raster. A raster to polygon conversion creates a vector layer of the tree footprint from which any polygons smaller than 8 square metres are filtered. This filter removes any non-tree artefacts that are misclassified in the first step such as power lines or telephone poles. This also filters out trees that are not large enough to provide shading benefits to citizens.
The Hot Spot Getis-Ord Gi* analysis identifies the locations where there are many census tracts showing high heat risk. The Hot Spot analysis uses fixed distance band and Euclidean distance parameters. The groups of census tracts that are identified as vulnerable at the 99% confidence level are isolated and compared back to the map of tract vulnerabilities to find the worst tracts in the area. The Hot Spot tool identifies areas that have a significantly higher vulnerability than surrounding tracts. The tracts in the areas identified with Hot Spot analysis that are classified as high priority are the final results of the model and show where increased tree cover is needed most.