Data on chemical facilities and relevant populations will be compiled for this study, specifically: (1) chemical facilities within the nation's 132 metropolitan areas will be characterized based on their threat potential as defined in Section 550 of the DHS Appropriations Act of 2007, and (2) geographic extent of the risk-shed for each facility using a worst-case scenario of a chemical release. These empirical data will be integrated into theoretical models of expected behavioral responses to populations at risk in an effort to spatially model the likely responses within the nation's urban areas. A proof-of-concept for this approach will be tested in Charleston and Columbia, SC.
While the results at the national scale showed some regions bear a disproportionate burden of security threats, this is not the case for all the different types of threats. For example, "release threats" are concentrated in the eastern half of the nation, and along the west coast. The most concentrated areas are along the Texas coast, in southern Louisiana, and the industrialized Midwest. "Theft threat" chemicals are less concentrated overall, but have clusters in the Great Lakes region. "Sabotage threat" chemicals are found mostly in the southern tier of states.
The results of the more detailed study in South Carolina showed the differences in populations at risk as well as in their differing vulnerabilities. Larger population centers (e.g. Charleston, Columbia, Greenville, Spartanburg) and their surroundings coincide with high potential release threats and populations at risk. Charleston and Columbia have the highest potential risk areas, but they manifest themselves slightly differently. In the case of Charleston, higher exposure is more concentrated near the core of the metropolitan area, while in Columbia the highest exposure has a decentralized pattern, mainly showing up in the periphery (surrounding suburban or rural areas) rather than in the urbanized core. In considering the vulnerability of the exposed populations there are also geographic differences.
In Charleston, there is a spatial overlap of high-risk areas and large numbers of exposed populations. However, the vulnerability is muted for some as these populations have greater wealth, but they also have extreme poverty. In Columbia, the most exposed populations have low to moderate incomes, but are not in poverty. Using a spatial knowledge discovery approach enables new hypothesis generation about risk factors and assists probabilistic risk assessment by pointing out the geographic and multivariate dimensions of risk (exposure plus sensitivity) factors. There is a wide range of socio-economic and demographic variables that define characteristics of populations at risk. High potential release threat, availability (location in relation) of health emergency response teams, and percent poverty had the most interesting patterns. Because of the data availability (only publically-available data were used), this study shows simplistic but promising results for the identification of geographic distribution of security risks, populations at risk, and their differing vulnerabilities. Further analysis is needed to find ways to incorporate exploratory findings into quantitative risk assessment models.
Two research questions provided the focus for this work:
(1) What is the regional variability in the potential threat of chemicals? and (2) What differences exist in the populations at risk overall, and in their differing vulnerabilities, based on releases, theft, and sabotage. To address these, the project used a combination of methods.
The first method was constructing the spatial data base. Data on Toxic Release Inventory (TRI) sites (for 2009) were used to determine the locations and the chemicals found in industrial facilities in the United States. Those chemicals and/or facilities in the TRI file that did not handle DHS chemicals of interest (COI) were eliminated. Spatial modeling of point source chemical hazards provided the spatial extent of the likely exposure surface (sabotage, theft, accidental release) as well as the populations at risk. The multivariate patterns of social vulnerability were examined using an integrated geographic knowledge discovery environment consisting of a self-organizing map and a parallel coordinate plot. Further analysis was conducted using multivariate spatial and non-spatial statistics and geo-visualizations. Finally, a case study of the methodology and results on the potential release threats in South Carolina was completed.