Develop a GIS Model to Identify Areas of Greatest Susceptibility to Heat-Related Illness
The second objective is to develop a GIS model to identify areas of greatest susceptibility to heat-related illness. This model overlays three types of vulnerability data and compares vulnerability between tracts. Figure 2 outlines the model, which runs in three segments. All data needs and sources are available in Appendix. First, LiDAR and imagery data are used to assess percent tree coverage in the City of Montreal. A tree canopy footprint layer is derived from the LiDAR point cloud. Tree-cover is aggregated to the Montreal census tracts, and the 25% of tracts with least tree coverage are identified for use in determining the highest-priority heat-vulnerable tracts. These tracts are assigned a tree-cover vulnerability score of 1.
Census data relating to heat vulnerability factors 1-7 are joined to the Montreal census tracts. In order to statistically compare socioeconomic vulnerability between tracts, percent prevalence of each census-derived socioeconomic variable is assigned to each census tract, following the method of Reid et al. (2005). Table 1 relates each vulnerability factor to its associated model metric. In addition to census data, a Euclidean distance analysis determines proximity to the nearest publicly accessible cooling center and assigns an average proximity value to each tract (Aminipouri et al., 2016).
An indexing system is applied to vulnerability variables 1-8. Following the method of Aminipouri et al. (2016), each vulnerability variable is unweighted. For each variable, the 25% of census tracts with the greatest percent prevalence of the associated vulnerability metric are identified as single-variable priority tracts. These tracts are assigned a single-variable priority value of 1. Each Montreal census tract is indexed based on the number of variables for which it is considered a priority tract, such that the highest multi-variable index score indicates the most vulnerable tract.
Overall tract vulnerability is derived from a sum of the multi-variable socioeconomic vulnerability index and the tree-cover vulnerability score. This yields a final vulnerability layer. This combined layer identifies high-priority tracts where both minimal tree-cover and prevalence of socioeconomic variables cause tract-level heat vulnerability.
The described indexing method is applied as an unweighted analysis of multiple criteria. Previous spatial analyses of socioeconomic heat vulnerability factors almost exclusively use an unweighted MCE (Ho, Knudby and Huang, 2015; Buscail, Upegui and Viel, 2012; Rebetez and Rong, 2005). Some identified socioeconomic variables may influence vulnerability more than others, but no empirical data supports an unequal weighting (Ho, Knudby and Huang, 2015). Indexing is more suitable for use by a municipality for tree-cover planning than an unweighted MCE, as it is reproducible by professional planners, it uses simple metrics (e.g. the 25% most vulnerable tracts), and its application does not require extensive geospatial analysis training.
Finally, a hot-spot analysis identifies regions larger than individual census tracts where there is significant need for increased tree coverage. The hot-spot tool calculates the Getis-Ord Gi* statistics for the features in the final vulnerability dataset (Mitchell, 2005). The result of this analysis provides z-scores and p-values, which indicate where there are high- or low-value spatial clusters, and locates areas of statistically significant vulnerability (Alessa, Kliskey and Brown, 2008; Rinner et al., 2010). The hot-spot tool is particularly useful in analysing coupled social and ecological factors.