Space and Time—Impacts on Modeling Extreme Heat Warning Systems

Heat is a leading cause of weather mortality throughout the world. Advanced warning and mitigation practices need to be researched and implemented to improve mitigation practices, particularly for urban environments. Previously, research has focused on identifying specific socioeconomic or environmental variables which are characteristic among victims of extreme heat events. The identification of these specific ‘predictive variables' can assist mitigation practices by identifying populations in which they occur. However, less research has been conducted on the interrelationship between Earth observation derived environmental and social variables, or whether these variables are resilient to aggregation error and time lapse. Following the methods of the previously developed extreme heat vulnerability index (EHVI), Principal Component Analysis was conducted on twenty-five documented indicators of heat-health hazards. Statistical results, including explanation of variance and correlation, were used to compare explanation of heat mortality distribution across varying spatial resolutions (US Census Tract, Block Group, and Block boundaries). Larger census tract boundaries demonstrated a higher percent of explanation among the heat death variability, but smaller boundaries had fewer components suggesting better variable correlation. This implies finer boundary models could improve vulnerability assessment if large scale environmental variables, currently collected by Landsat imagery, can be incorporated at a comparable resolution. The coarse resolution environmental data is more prone to aggregation bias and can hinder vulnerability modeling within smaller boundaries. The procedures were replicated to include a time lag to identify whether the previously identified indicators of vulnerability would provide consistent results over time, or if changes in population demographic would impact mitigation. Initial results suggest small scale analysis (such as block group boundaries) provides improved mitigation potential over currently implemented weather service warnings, and population changes over time alter vulnerable population patterns.