Climate change has increased the frequency and intensity of extremely hot weather. The health risks are not uniform across affected areas due to variability in heat exposure and social vulnerability, but these differences are challenging to map with precision. Therefore, this PhD research evaluated the uses of linear and nonlinear statistical models to map the intra-urban difference of temperature distribution, compared the temperature distributions mapped from land surface temperature, air temperature, and apparent temperature (Humidex), and developed a spatially- and temporally-stratified case-crossover approach for delineation of areas with higher and lower risk of mortality on extremely hot days, and applied this framework in greater Vancouver, Canada. A digital elevation model, Landsat 5 TM, Landsat 7, and MODIS water vapor products were used to map the air temperature and Humidex. Records of all deaths with an extremely hot day as a case day or a control day were extracted from an administrative vital statistics database spanning the years of 1998-2014. Three heat exposure and eleven social vulnerability variables were assigned at the residential location of each decedent. Conditional logistic regression was used to estimates the odds ratio for a 1°C increase in daily mean temperature from a fixed site for subsets of the data above and below different values of the spatial variables. An important result was that intra-urban variability in air temperature and Humidex could be mapped using a Random Forest model with good accuracy (RMSE = 2.31°C and 2.04°C respectively). This project also found that Humidex could better demonstrate the spatial distributions of heat exposure when it was compared with the temperature maps from land surface temperature and air temperature. Also, the heat exposure and social vulnerability variables with the strongest spatially-stratified results were Humidex and labour nonparticipation rate. Areas at higher risk had values ≥3.7°C warmer than Vancouver International Airport, the reference site, and ≥60% of the population neither employed nor looking for work. These variables were combined in a composite index to quantify their interaction and enhance visualization of high-risk areas. In conclusion, methods from this PhD research provided a data-driven framework for spatial delineation of the temperature-mortality relationship by heat exposure and social vulnerability. Results can be used to map and target the most vulnerable areas for public health intervention, and the methodology is directly transferable to other cities in Canada and abroad.
Copyright is held by the author.
This thesis may be printed or downloaded for non-commercial research and scholarly purposes.
Member of collection