Comparing the Impact of Spatial Resolution on Modeling Extreme Heat Vulnerability

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015: 2:00 PM
226AB (Phoenix Convention Center - West and North Buildings)
Austin C. Stanforth, Indiana University, Indianapolis, IN; and D. P. Johnson

Heat is currently considered the leading cause of weather related fatality throughout the world. Researchers have primarily focused on identifying socioeconomic or environmental variables which predict vulnerability, or level of risk, during an impending extreme heat event. More research is needed on the interaction of these variables at diverse spatial resolutions. Variables can interact differently at varying spatial boundaries. Understanding how an aggregation boundary impacts a variable, or which variables are more prone to this bias, is important to the future of vulnerability modeling at finer scales. Following the methodology of the previously developed extreme heat vulnerability index (EHVI), Principal Component Analysis was conducted on twenty-five well documented indicators of heat-health hazards. The analysis was conducted at varying spatial resolutions (US Census Tract, Block Group, and Block boundaries) and the results were compared by their explanation of variance and correlation data. The results showed the census tract boundary explained a higher percent of the heat death variability, but smaller boundaries had fewer components suggesting better variable correlation within the model. This implies finer boundary models can improve vulnerability assessment. However, large resolution environmental variables, collected through Landsat imagery, contributed less during smaller boundary tests. This suggests coarse resolution environmental data is more prone to aggregation bias and can hinder vulnerability modeling within smaller boundaries. The results support previous studies which identified individual variables, such as age and education, to be more predictive of vulnerability and less prone to aggregation bias. This study demonstrates vulnerability models using smaller boundaries are more effective than the county wide warning systems currently used by the National Weather Service. Implementation of finer models could develop improved mitigation strategies, reduce heat mortality, and decrease medical expenses during extreme heat events.