Natural Hazards Mortality in the United States
Advisor: Dr. Susan L. Cutter
Hazards research often focuses on either pre-event concepts such as societal vulnerability, or post-event impacts such as hazard mortality. It is necessary to study each of these concepts individually, but researchers must remember that pre- and postevent concepts are dynamically linked and influence one another. Despite the dynamic relationship between pre-event vulnerability and post-event outcomes, few conceptual models allow for integration before and after a hazard event. Therefore, theoretical and empirical research that incorporates the relationship between vulnerability and impacts is lacking.
Two concepts, social vulnerability and mortality from natural hazards were examined in this dissertation as pre- and post-event concepts. This research was conducted within a vulnerability framework and expanded a well-known theoretical model to include post-event impacts. This provided a theoretical base from which social vulnerability and hazard mortality could be empirically tested. However, spatial trends in hazard mortality are not well documented, largely due to data availability constraints. Before investigating the relationship between social vulnerability and hazard mortality, a hazard fatality database was constructed and explored. Furthermore, confounding variables that could influence the relationship between social vulnerability and hazard mortality were identified and included in the analysis. Therefore, the questions that guided this research were:
1. What spatial and temporal patterns are evident in hazards-induced mortality?
a. Are areas of abnormally high and low mortality clustered in space?
b. Do the spatial patterns of mortality demonstrate spatial persistence over time?
2. What is the relationship between hazard mortality and selected covariates?
3. Is there a spatial and/or statistical relationship between pre-event social vulnerability and hazard mortality as an outcome?
Various measures of mortality were included in the analysis, and each displayed distinct spatial patterns. The two most prominent metrics were crude mortality rates (CMRs) and standardized mortality ratios (SMRs). The former showed the impact of hazard mortality over time, and revealed spatial patterns that included low CMRs in urban centers and larger impacts in the interior United States. By contrast, SMRs demonstrated higher risk of death in coastal and urban locations, but low risk throughout the central U.S. Both measures displayed statistically significant local clusters over time and space.
A review of relevant literature led to selecting three variables as covariates to hazard mortality: seasonality, the urban nature of counties, and counties’ propensity for extreme events. Few variables demonstrated high explanatory power with respect to hazard mortality. An exception was the urban variable in explaining risk of death through SMRs. However, all variables showed significant spatial variation in local regression analyses, indicating that the relationships between hazard mortality and covariates are not constant across space. A bivariate cluster analysis, performed as a preliminary assessment of the relationship between social vulnerability and hazard mortality, revealed a random spatial relationship. Holding the three covariates constant in a statistical analysis did not improve these results. Social vulnerability demonstrated no relationship with hazard mortality. Like the covariate analysis, a local regression procedure revealed spatial variation in the relationship. This indicates that although the relationship between social vulnerability and hazard mortality is local and might not be best suited for national level inquiry.
This dissertation demonstrated a conceptual/theoretical link between social vulnerability and hazard mortality. This relationship was tested empirically using spatial statistical analysis. The research in this dissertation represents a first attempt at incorporating hazard impacts into a vulnerability context. The results from the empirical tests did not support the conceptual relationship identified in relevant bodies of literature. However, these results indicate that the relationship between social vulnerability and hazard mortality is local. Therefore, this dissertation is useful for site/regional selection of local areas of interest. Case studies can then be performed in these areas to further understand why the expected relationship between social vulnerability and hazard mortality holds true in some areas but not others. Further refinement of social indicators and hazard mortality data is necessary for more reliable results that could be used by decision makers and emergency managers to reduce both vulnerabilities and impacts to natural hazards.Dissertation