Model-Based Clustering of Social Vulnerability to Urban Extreme Heat Events

摘要

Geodemographic classification methods are applied to Denver Colorado to develop a typology of social vulnerability to heat exposure. Environmental hazards are known to exhibit biophysical variations (e.g., land cover and housing characteristics) and social variations (e.g., demographic and economic adaptations to heat mitigation). Geodemographic model-based classification permits a more extensive set of input variables, with richer attributions; and it can account for spatial context on variable interactions. Additionally, it generates comparative assessments of environmental stress on multiple demographic groups. The paper emphasizes performance of model-based clustering in geodemographic analysis, describing two stages of classification analysis. In so doing, this research examines ways in which high heat exposure intersects with socioecological variation to drive social vulnerability during extreme heat events. The first stage classifies tract-level variables for social and biophysical stressors. Membership probabilities from the initial (baseline) classification are then input to a second classification that integrates the biophysical and social domains within a membership probability space to form a final place typology. Final place categories are compared to three broad land surface temperature (LST) regimes derived from simple clustering of mean daytime and nighttime land surface temperatures. The results point to several broad considerations for heat mitigation planning that are aligned with extant research on urban heat vulnerability. However, the relative coarseness of the classification structure also reveals a need for further investigation of the internal structure of each class, as well as aggregation effects, in future studies.