ABSTRACT. Political leaders continue to reaffirm citizens' right to know about possible health hazards from industrial facilities located in their neighborhood. Yet, due to the complexity of conventional risk characterizations, it is extremely difficult for the average citizen-activist to discern the relative risk posed by local hazard sources. This paper describes the development of a comparative measure of risk for industrial facilities in South Carolina. Using Toxic Release Inventory (TRI) data, the paper provides a methodology for conducting comparative risk assessments for individual facilities and communicating these risks to the general public. The risk visualization tool was developed using geographic information systems (GIS) and digital cartography. The prototype (South Carolina Toxic Risk Atlas) is a static web-based tool that is now in a testing and feedback stage (http://www.cla.sc.edu.pallas2.tcl.sc.edu/geog/hrl/sctrap/index.htm).
Recently, the U.S. Environmental Protection Agency (USEPA) announced a plan to provide pollution profiles of selected industries (automobile, steel, metals, oil refining, and papermaking) via the Internet (Cushman 1997). These profiles would include releases and spills, pollution-to-production ratios, as well as data on inspections, non-compliance, and enforcement actions. In addition, the USEPA plans on providing a hazard indicator for each plant based on the relative toxicity of the chemicals released and some demographic data on residents within three miles of the plant.
The issue of deriving a relative measure of risk to compare individual facilities or different types of contaminants is not new and has been part of USEPA's agenda for many years. The risk assessment paradigm (describing the detrimental effects of industrial toxins on people and ecosystems that support human activity) has dominated most of these discussions. The methods used to conduct risk assessments include four steps: hazard identification, dose-response assessment, exposure assessment, and risk characterization (National Research Council 1983, 1994). However, as Nyerges et al. (1997) discussed earlier, the practices of risk assessment are coming under attack. The primary points of contention include data needs, data compatibility, and data uncertainty (National Research Council 1994), the nature of acceptable risks and acceptable evidences (Mayo and Hollander 1991, Kunreuther and Slovic 1996), prioritizing and comparing risks (Davies 1996), and communication of technical results to decision makers and the public (NRC 1996).
A central issue in characterizing and communicating risks is the representation of those risks graphically and in an accessible manner. Risk information varies across space, and the variability in risk perception is partially dependent on the quality of data received. GIS and digital cartography will play a large role in risk assessment and communication in the future as we move away from strict deterministic models of risk to more socially constructed approaches to risk assessment. This shift was recognized by the National Center for Geographic Information and Analysis (NCGIA) as part of Initiative 19--GIS and Society. Central to this initiative are key questions on how GIS influences the perception of risk, how GIS can empower or disempower community groups, and how equity issues can be examined within a socio-spatial context (Harris and Wiener 1996).
Related to the empowerment question is the analysis of the disproportionate distribution of risk on people and places. The idea of environmental inequities has captured the attention of many activists and researchers following the publication of the watershed study by the United Church of Christ (1987) and Bullard's Dumping in Dixie (1990). Concern about the correlation between race and hazardous facility location has prompted a surfeit of research on environmental justice topics (Cutter 1995). Differences in the environmental threat examined, the scale of measurement, the subpopulations sampled, and the time frames have led to ambiguous conclusions in the empirical support for or against environmental (in)justice.
This paper describes the development of a comparative measure of environmental risk in order to contrast facilities and spatial units at the state and local levels. Our primary purpose is twofold: 1) to illustrate a simplified methodology for conducting comparative risk assessments; and 2)
provide a tool for risk visualization and communication to the public, using a GIS and digital cartography framework. We contend that to develop relative risk measurements at the community level, one must be cognizant of the accuracy of the database, the differences in the magnitude and toxicity of industrial toxins, the transport and fate of these toxins, and the degree to which community activists, business leaders, and decision makers understand risk visualizations.
The Context of Comparative Risk
Since the publication of USEPA's Unfinished Business (1987), locally based risk assessments are burgeoning. Through USEPA support, more than three dozen regions, states, and municipalities have conducted some form of comparative risk study (Minard 1996). USEPA has also established its Community-Based Environmental Protection (CBEP) program in an effort to support local risk projects (Wernick 1996), realizing that national comparisons often mask locally important risks. While providing a good baseline of data, most of these efforts only rank risks along a variety of dimensions; they do not spatially represent them (Wernick 1997).
Correlating Risks: The Spatial Domain
Some of the earliest comparative risk projects were conducted in the 1970s. Berry (1977), for example, mapped pollution indicators (water, air, solid waste, noise) for 13 metropolitan areas in an effort to understand the social burdens of environmental pollution. More recently, Goldman (1991) examined the relationship between demographic characteristics, health outcomes, and multiple risks (industrial toxins, pesticides, water quality, air quality). Using county level data, Goldman depicted regional variation in risk exposure and mortality.
Using multiple indicators of toxic releases, Cutter and Solecki (1996) compared the spatial distribution of airborne toxic releases by county in the Southeast in an effort to define the relative hazardousness of places based on these acute and chronic emissions. They also questioned whether lower-income, minority counties were disproportionately at risk and found no conclusive evidence of socio-demographic inequities. In another example using South Carolina (Cutter et al. 1996), three indicators of industrial toxins were mapped at three different spatial scales. There were distinct geographic variations in risks across the state, although the pattern of socio-demographic impacts was not as well defined. Finally, Stockwell et al. (1993) utilized GIS to delineate the spatial patterning of potential risk exposures from TRI facilities for the Southeast. Utilizing magnitude estimators for this single risk indicator, they found the largest TRI releases were near the most densely populated areas.
Contrasting Impacts: The Social Domain
Another approach to comparative risk assessment is to take a single risk indicator (such as Superfund sites or Toxic Release Inventory facilities) and examine potential exposures of areas based on their socio-demographic profiles. These studies provide empirical data on the social inequalities of potential impacts. However, much of this research is more statistically than spatially oriented (see Cutter 1995, Cutter et al. 1996 for a review). The use of GIS in illuminating spatial equity has been demonstrated in a number of recent studies (Burke 1993, Bowen et al. 1995, Glickman 1994, Glickman et al. 1995, McMaster et al. 1997, and Sui and Giardino 1995).
In this paper, we have opted to concentrate on comparing risks and communicating these to the public, rather than providing a spatial equity study (see the remaining papers in this volume). However, it is an easy translation to move from the delineation and communication of relative risk to the incorporation of social and demographic indicators of potentially affected populations.
Data for this research were taken from three of USEPA's national databases on hazardous waste and toxic releases: CERCLA's National Priorities List (NPL); EPCRA's Toxic Release Inventory (TRI); and RCRA's Biennial Reporting System (BRS). For more detailed descriptions of these databases see Scott et al. (1997a). Our paper utilizes 1992 data because at the time this research was launched (1996), these data were the most current. It is important to be cognizant, however, that the data for each of these hazard indicators varies year-to-year because of changes in reporting releases and in the listing and/or delisting of facilities.
Developing the Hazard Source Inventory
When developing a catalog of hazardous facilities, it is erroneous to simply tabulate records in separate databases because some facilities appear in several databases due to reporting requirements. Moreover, single facilities can and do produce multiple exposures. The USEPA recognizes this and assigns a unique facility identifier in the TRI, BRS, and CERCLIS databases. Unfortunately, the USEPA identification number did not prove to be unique in all cases. We therefore developed a procedure for creating a unique identifier or key field (Scott et al. 1997a). In this way, we were able to insure a many-to-one correspondence between South Carolina facility listings and their characteristics across the three databases. Finally, the positions of many of the facilities were incorrect. Procedures for error estimation and correction protocols are reported elsewhere (Scott et al. 1997b).
Not only is the locational accuracy of hazardous facilities an important consideration in comparative risk assessments, so is an estimate of the magnitude of potential chemical exposure from those facilities. Unfortunately, there are some inherent obstacles in the chemical information reported in the databases that, at this time, prevent us from accurately estimating hazard magnitudes from all our risk sources. Specifically, RCRA's Biennial Reporting System summarizes the hazardous waste generated and treated not by chemical but by the generation and treatment process. This means that the majority of wastes are reported as "soups" consisting of several different, often unknown, chemicals. Determining the hazardousness of CERCLIS sites is not possible because most of the sites in our state have not had a preliminary assessment or site inspection. The National Priority List sites may have had land and water samples taken and tested for the concentration of chemicals, but the actual amounts of toxins present at a site are not known.
Given the problems in the chemical information found in the BRS, CERCLIS, and NPL databases, we concentrated on TRI data to estimate the hazardousness of industrial facilities. When creating a magnitude estimator, it is important to take into account both the magnitude (amount of chemicals released) and the toxicity of the release. Fortunately, the magnitude of the TRI releases is standardized to pounds per year. Unfortunately, there is no universally accepted measure of environmental hazardousness, largely due to insufficient toxicological studies on the behavior of many chemicals in a natural setting. While there has been a plethora of research on the toxicity of chemicals in industrial or workplace settings, data on many of the more than 600 regulated chemicals' properties in the open environment are still sparse. Therefore, we borrowed the industrial hygiene measurements of toxicity as our relative measurement of environmental toxicity (ACGIH 1991). Our reasoning is that while the exact levels of dangerous exposure are bound to change in an outdoor environmental setting, the relative hazardousness of a chemical should remain constant. Our relative measure is based on the Threshold Limit Value -- Time-Weighted Average (TLV-TWA) or the maximum average amount in mg/[m.sup.3] to which a worker can be exposed in a given eight-hour workday. These TLVs were converted to their inverse values (1/TLV), so that the greater the number, the higher the relative toxicity.
While we feel that using TLVs as a relative measure of toxicity is reasonable, it is important to point out a few caveats. First, not all of the chemicals reported by TRI facilities in 1992 have TLVs, although the vast majority (89%) do. Second, the medium of exposure often determines the level of toxicity, but we have assigned one number per chemical, regardless of physical state (vapor, liquid, solid). Fortunately, most of the releases we classified with the environmental TLVs are airborne, the medium on which most occupational TLVs are based. Finally, it has been suggested that chemicals with high TLVs may not be more toxic than others, just more closely studied. There is little we can do to remedy this potential bias, except to echo calls for more research into the toxicity of industrial and environmental toxins. Incidentally, Bowen et al. (1995) reached the same conclusion, adopting threshold limit values as their measure of toxicity, however their creation of a composite toxicity index was significantly different than the weighted average toxicity index described here.
Once a measure of relative toxicity for each chemical was created, there was the need to relate that toxicity to a given facility. However, because most facilities release more than one chemical, a system for combining several toxicity values was required. We created an index called a weighted average toxicity (WAT) value (Appendix). The WAT allows us to construct a measure of relative risk when comparing two facilities in terms of the magnitude and toxicity of their releases. For example, the WAT helps differentiate low emissions of highly toxic substances from high emissions of relatively non-toxic ones. In this way, we are able to move beyond the quantity of releases, which is what is routinely reported to the USEPA, to a more realistic assessment of the risk which includes both quantity released and toxicity.
To calculate a WAT value for a given facility, we summed the total number of pounds of chemicals the facility releases in a year. Second, we found what proportion each chemical adds to the total (i.e., if the total pounds equals 10,000, then releasing 1,000 pounds of a given chemical would represent 10% of the total). Third, we multiplied each proportion by the relative toxicity of respective chemicals. Finally, we added the products together to get a weighted average toxicity for each facility for a given year. The mathematical notation used is:
(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
[WAT.sub.f] = weighted average toxicity value for a given facility
n = number of chemicals released by a given facility
[C.sub.i] = amount of chemical i (in pounds)
[C.sub.t] = total amount of chemicals released (in pounds)
[TLV.sub.i] = threshold limit value of chemical i.
Table 1 gives an example of the computation of the WAT for two different facilities in Lexington County (part of the Columbia metropolitan area). Once calculated, the weighted average toxicity (WAT) for each facility can be compared to others in the county or state. For example, the WATs could be ranked from most toxic to least (Table 2). For the 1992 TRI facilities, the WAT values range from 0.0001 to 100.0.
Table 1. Reported releases and computation of weighted average toxicity. Example 1: Cooper Power Tools. 0.54 x 0.0037 = 0.0020 0.46 x 0.0038 = 0.0017 WAT = 0.0020 + 0.0017 = 0.0037 Chemicals Pounds Released % Total 1/TLV Trichloroethylene 1 300 0.54 0.0037 Methanol 1 100 0.46 0.0038 Example 2: Owen Electric Steel Company 0.070 x 2 = 0.0140 0.3699 x 0.2 = 0.0740 0.0699 x 6.67 = 0.4662 0.0077 x 5 = 0.0385 0.5434 x 100 = 54.34 0.0005 x 1 = 0.0005 0.0014 x 2 = 0.0028 WAT = 0.0140 + 0.0740 + 0.4662 + 0.0385 + 54.34 + 0.0005 + 0.0028 = 54.94 Chemicals Pounds Released % Total 1/TLV Chromium compounds 851 0.0070 2.00 Manganese compounds 45 078 0.3699 0.20 Lead 8 522 0.0699 6.67 Copper 938 0.0077 5.00 Zinc (fume or dust) 66 250 0.5434 100.00 Nickel 63 0.0005 100 Barium 172 0.0014 2.00
Table 2. Ranking of most toxic TRI facilities based on weighted average toxicity in 1992.
Rank Facility Name City County 1 Nucor Steel Co. Darlington Darlington 2 Profession Medical Products Co. Greenwood Greenwood 3 ACM Corp. Summerville Dorchester 4 Georgetown Steel Co. Georgetown Georgetown 5 Owen Electric Steel Co. Cayce Lexington 6 Southern States Galvanizing Co. Traveller's Greenville Rest 7 B.L. Montague Co. Sumter Sumter 8 Gaston Copper Recycling Corp. Gaston Lexington 9 Hitachi Electronic Devices Greenville Greenville 10 AVM Inc. Marion Marion Number of Amount Majority Majority Rank Chemicals (lb.) WAT Chemical % of Total 1 1 22 438 100.00 Zinc 100 2 2 5 650 95.58 Zinc 96 3 2 4 203 82.12 Zinc 82 4 4 29 589 68.63 Zinc 68 5 7 21 874 54.94 Zinc 54 6 3 2 510 47.92 Zinc 48 7 3 22 862 45.94 Methyl ethyl 53 Ketone 8 7 108 855 40.69 Zinc 37 9 7 72 698 24.14 Lead 47 10 5 2 500 20.83 Chromium 20
In and of itself, WAT rankings provide only limited information. Adding information on the amount released, the number of different chemicals released, and the dominant chemical as a percentage of the total releases provides a more complete profile of the facility. Facilities can be differentiated based on high emissions of lower toxicity (HE/LT), such as Hitachi Electronic Devices, from those with lower emissions but higher toxicity (LE/HT), such as Professional Medical Products (Table 2).
The majority of high-WAT facilities emit heavy metals such as zinc, lead, and chromium. However, if one ranks by total emissions, a different listing is produced (Table 3). This ranking shows that the state's largest emitters which (according to Livingston 1996) have consistently been listed as the worst in the state, release relatively low-toxicity substances such as methanol. The examination of Tables 2 and 3 in tandem illustrates the high magnitude/low toxicity (HE/LT) and low magnitude/high toxicity (LE/HT) continuum. The dichotomy points to the critical need to examine both magnitude and toxicity indicators in any comparative risk assessment for individual facilities or county-wide comparisons. Once we created the relative risk components, we explored ways to represent the risk graphically and to make these visualizations readily available to the general public.
Table 3. Ranking of most toxic TRI facilities based on total pounds released in 1992.
Rank Facility Name City County 1 Westvaco W. Charleston Charleston 2 Westnghouse Electric Co. Hampton Hampton 3 Bowater Inc. Catwba York 4 Stone Container Corp. Florence Florence 5 Hoechst-Celanese Corp. Rock Hill York 6 Anchor Continental Columbia Richland 7 Carolina Eastman Columbia Lexington 8 Nicca USA Inc. Fountain Inn Laurens 9 International Paper Inc. Georgetown Georgetown 10 Teepak Inc. Swansea Calhoun Number of Amount Majority Majority Rank Chemicals (lb.) WAT Chemical % of Total 1 14 5 353 180 0.11 Methanol 77 2 11 4 756 120 0.01 Methanol 84 3 14 3 948 992 0.28 Methanol 52 4 6 3 044 712 0.08 Methanol 47 5 19 2 965 008 0.037 Acetone 76 6 2 2 948 702 0.012 Toluene 99 7 16 2 888 153 0.222 Acetaldehyde 25 8 2 2 235 790 0.999 Ammonia 96 9 10 2 060 155 0.055 Methanol 65 10 2 1 708 000 0.085 Ammonia 97
The visualization of risk was the most challenging part of this research. Magnitude and toxicity maps using graduated circles were produced for the entire state. We then created a series of maps showing pounds released classified by weighted average toxicity. In this way, we present the interaction between magnitudes and toxicities at the individual facility level and how this varies from site to site across the state.
Figure 1 illustrates this visualization using 1992 data. In South Carolina, the majority of facilities (241) fall into the lowest toxicity class and have moderate amounts of releases (Fig. 1a). Most of these facilities are dustered in the upstate region, the most industrialized portion of the state. Facilities with the largest emissions generally have either low or moderate toxicities (Fig. 1b). They tend to be located in smaller mill towns throughout the state and in the state's major metropolitan centers. Highly toxic emissions are released by only a few facilities (89), but they normally are at very low quantities (Fig. 1c). The one exception is the Union Camp paper/pulp mill in lower Richland County, which had a total emission of 1,324,725 pounds in 1992 and a WAT of 1.48. The emissions (by percentages of the total) were methanol (57%), hydrochloric acid (28%), and sulfuric acid (5%), as well as several other chemicals released in small amounts.
[Figure 1 ILLUSTRATION OMITTED]
Moving beyond a composite facility-based indicator, we next attempted to visualize risk from an aggregate perspective by comparing all 46 counties within the state. Counties were classified by WAT and total pounds released and placed into a 3 x 3 matrix (Figure 2). As can be seen, there were no counties with large releases in the medium to high toxicity range. The riskiest counties had moderate to low amounts of emissions with high toxicity (LE/HT) or high amounts of emissions with low toxicity (HE/LT). With one exception, all of these counties are metropolitan areas adjacent to the state's urban centers in Charleston, Columbia, Greenville, and Rock Hill, SC-Charlotte, NC. The one anomaly is Westinghouse Electric located in rural Hampton county in the southern portion of the state along the South Carolina/Georgia border.
[Figure 2 ILLUSTRATION OMITTED]
Locally relevant information on risks was provided as well. This involved presenting the toxicity and magnitude indicators at a more detailed scale such as census enumeration districts. One example is Lexington County, located in the middle of the state (Figure 3). Lexington County was selected because it has moderate levels of emissions, yet these are rather high in toxicity. Mapping both the pounds released (Figure 3a) and weighted average toxicity (Figure 3b) furnished an interesting pattern at the local level.
[Figure 3 ILLUSTRATION OMITTED]
The majority of releases in the county come from one plant, Carolina Eastman (2.9 million pounds in 1992), but the WAT of the emissions is rather low (0.22). In contrast, the highest toxicity releases come from the Owens Electric Steel and the Gaston Copper plant (now defunct), even though the quantities emitted were less than 5% of the total releases from Carolina Eastman. Most of the emissions at the Owens Steel plant are from electroplating processes involving the release of heavy metals such as zinc (Table 1) with an extremely high WAT. Gaston Copper's high relative toxicity was also due to the release of heavy metals, but from a recycling process.
Finally, we experimented with a technique for visualizing the risks which involves representing individual chemicals and their toxicity by facility. The technique is illustrated in Figure 4, where the areas of the circles represent the total pounds released, while the divisions signify what proportion different chemicals contributed to the total pounds released. The shading of the pie-chart wedges indicates each chemical's relative toxicity.
[Figure 4 ILLUSTRATION OMITTED]
It is difficult to compare chemicals from facility to facility because each uses different manufacturing processes resulting in different products and chemical releases. However, we can depict how important (in terms of quantity released) and how toxic a contributing chemical is in the overall computation of the WAT. This generalization can then be portrayed spatially to illustrate the differing hazard profiles of facilities within a community or county.
Figure 4 serves to illustrate this representation for Lexington County. One can see that while Carolina Eastman released almost 3 million pounds of chemicals in 1992, almost 75% had a relatively low toxicity. This plant can be compared to Gaston Copper or Owen Electric Steel which released highly toxic chemicals in relatively small amounts. This type of visualization allows one to move beyond one or two numbers representing hazardousness and begin to extract the complex and conflicting information about magnitudes and toxicities for given facilities.
Communicating Risks to the Public
Following the database creation and magnitude analysis, a means of communicating these risks to the public was sought. We were concerned about how to communicate a complicated set of data to a general public not familiar with evaluating toxic risk. Our primary consideration was to provide the information to as many people as possible in a form that would be easily understood. A second priority was to locate a computer system for the visual representation of the information. Libraries are a natural location for such information but not all libraries have the same computer capabilities, such as graphical user interfaces or Internet access.
A vital decision in communicating any information to the public, particularly in digital form, is the selection of an appropriate data/user interface. In selecting a given interface, system designers can either empower users with unprecedented possibilities for understanding, or they can channel user interaction so that a limited amount of information can be retrieved (Kuhn 1992). An interface can overwhelm a user with exploration possibilities or release information in manageable amounts, thus enhancing understanding and facilitating the risk communication process.
We had several design goals for our risk communication system. First, it had to provide text, maps, and graphs at a level suitable for general public consumption. Second, the interface had to be easy to learn with minimal expert interaction. Third, a level of interactivity was necessary, given that the communication of personal neighborhood information was the objective. Fourth, it had to be inexpensive or free so that wide dissemination throughout the state would be possible. Finally, it had to be accessible to most computer users, regardless of operating systemor hardware capabilities.
When we were searching for a suitable risk communication system in early 1996), several options were available to us. First, a customized user interface could be created from scratch, using a high-level programming language. While this would provide the maximum amount of control over the interface properties and capabilities, the cost of development was too high. Second, we could use an existing mapping package, such as ArcView 2.1. The interface can be customized to limit user choices while retaining a high level of interactivity. However, this option also posed problems since it is not free, not completely cross-platform-compatible, nor is it compact.
We concluded that a web browser interface (such as Netscape Navigator 2.0[TM)] would be the most accessible and best understood by all levels of potential users. The web browser allows easy point-and-click access to the information within the system via its hypertext conceptual model. In addition, this browser and data can be compacted onto disks and transferred to another personal computer. The software was free, meaning that the disks could be distributed to libraries and others interested in hazards assessment. Unfortunately, a level of interactivity was lost when moving to this solution, because creating dynamic map displays was extremely difficult at that time.
South Carolina Toxic Risk Atlas Prototype (SCTRAP)
The first step in system development was to create a prototype. Our goal regarding the prototype was to explore the various options for revealing risks to the community. This involved interactive maps, tables, lists, static maps, text, and hypertext. In the beginning of the development processes, we realized that the audience would need some general knowledge about environmental risks and hazards, so a fair amount of background information was encoded and made available to the user on demand. This included prototype goals, federal regulations, a glossary, and factsheets on the effects of chemicals on human health produced by the Agency for Toxic Substances and Disease Registry (ATSDR). To simplify construction, only one county in South Carolina--Richland--was used as a proof-of-concept.
The second and most time-consuming portion of prototype construction was the creation of interactive maps. The word "interactive" is used loosely here because while the user can choose which neighborhood maps to view, the scale, positioning, and the information presented on the maps is fixed. Before mapping the risk information, considerable discussion took place as to the nature of the ancillary information that would be provided to the user. Finally, it was decided that the maps would include easily recognized landmarks--schools, fire stations, interstate highways, U.S. highways, major water bodies, and incorporated area boundaries--as well as the sources of risk. Both the risk and the ancillary information was then visualized in ArcView 2.1 at three scales: the state (which shows only county name); the county (which shows only risk sources, interstate highways, water bodies, and incorporated areas); and the community level.
Each community-level map, of which there are 28 in Richland County, covers approximately 36 square miles. This size was chosen as a compromise between visualizing the neighborhood, the community, and the surrounding communities (Figure 5). The county index map is linked to the community-level maps through a common gateway interface (CGI) script called Imagemap. The Imagemap application allows the definition of an area of a GIF image as being linked to another HTML resource. Thus, when a user clicks on a given map square, he or she is taken to the page with that map.
[Figure 5 ILLUSTRATION OMITTED]
The third task in constructing our risk communication tool was to encode the risk information for individual facilities. This risk information included a picture of the facility (when available), the facility's address, the latitude/longitude position, and the emissions information given under different regulations (see above). The information varies by database. For example, TRI release listings contain the pounds released from the smokestack (or equivalent), the pounds released from fugitive sources, the chemical released, and the chemical's relative toxicity. In the prototype, each of the column headings as well as the chemical names themselves are linked via hypertext to the appropriate listings in the glossary or the ATSDR chemical fact sheets. The user can select the facility's symbol on the community level map, its name from the map legend, or its name from a text-based listing of all risk sources per county to access the individual facility pages.
The fourth and final piece of the prototype were the static maps of risk. In order to explore the public reaction to different types of risk visualization and to give users a comparative statewide perspective, we used static maps of magnitude and toxicity. These maps (Figure 1) were displayed separately from the community risk source maps to keep from confusing users; the community maps have three different sources of risk, the static maps show only TRI magnitude and toxicity.
One conclusion that can be drawn from this research is that while an indication of the amount of relative risk an industrial facility poses can be determined using total pounds released and a toxicity index, we are far from discussing the risk potential of environmental releases of toxic chemicals with any certainty. After decades of research, scientists still have little knowledge of the way chemicals affect living organisms in the environment. Between differences in transport media, varying chemical properties, lack of toxicological research, and the importance of individual's risk factors (age, weight, sex, etc.), a true measure of risk continues to be elusive. Unfortunately, we cannot wait for science to catch up with political reality. Every day, citizens and their leaders are making important decisions regarding personal and neighborhood exposure to toxic chemicals. Surrogate and incomplete measures of risk must not only be calculated, they must also be communicated to the public in an efficient and illuminating way.
The South Carolina Toxic Risk Atlas Prototype (SCTRAP) (http://www.da.sc.edu.pallas2.tcl.sc.edu/geog/hrl/sctrap/) represents an initial foray into the visualization and communication of toxic risks to the public. As is the case with all prototypes, we will change a number of the features when the system is implemented, using improved software and comments from users. First, we will expand and update the prototype to include all counties in the state as well as more recent data (1995 TRI, 1995 BRS, and 1997 CERCLA data). We will add other environmental risk data such as hazmat accidents; extremely hazardous and/or criteria air pollutants regulated under the Clean Air Act (i.e. ozone, sulfur dioxide, etc.); water quality indicators; and so forth.
The use of more and different visualizations based on user feedback would be beneficial. It has become clear that many of our users lack the geographical sophistication to adequately interpret many of the visualizations (graduated circle, choropleth maps) that we take for granted as geographers. In future versions, we will include more charts, graphs, and text descriptions to expedite map interpretation. Expanding the interactivity of the prototype is another goal. Recently released software can facilitate this interactivity (i.e. ESRI's Map Objects[TM]). Web users can now type addresses, perform queries, and manipulate visualizations on-line and in real time. This level of interactivity will help us solve some of the limitations identified by users.
The development of a risk visualization tool can assist local communities in making geographic comparisons of risk. It can also enhance access to and understanding of environmental data. The delivery mechanism supports the community right-to-know disclosure provisions of many environmental laws and will certainly help inform residents of risks in their community. Once armed with this information, it will be up to the residents and community leaders to decide on an acceptable level of risk from toxic chemicals in their community.
We would like to thank Charmel Menzel, Dan Wagner, Minhe Ji, and Lloyd Clark for their initial work on this project. Also thanks to Deborah S. K. Thomas, and Jerry T. Mitchell for contributing critical and editorial comments on this manuscript. Finally, we are grateful to the South Carolina Universities Research and Education Foundation (#30-95, 62-96) for providing funding for the risk atlas prototype.
American Conference of Governmental Industrial Hygienists (ACGIH). 1991. Documentation of the threshold limit values and biological exposure indices. Sixth ed. Cincinnati, Ohio: ACGIH.
Berry, B. J. L. 1977. The social burdens of environmental pollution. Lexington, Mass.: Ballinger.
Bowen, W. M., M. J. Sailing, ICE. Haynes, and E. J. Cyran. 1995. Toward environmental justice: Spatial equity in Ohio and Cleveland. Annals of the Association of American Geographers 85(4): 641-63.
Bullard, R. D. 1990. Dumping in Dixie: Race, class, and environmental quality. Boulder: Westview Press.
Burke, L 1993. Race and environmental equity: A geographic analysis in Los Angeles. Geo Info Systems October: 44-50.
Cushman, J. H. 1997. E.P.A. is pressing plant to publicize pollution data. The New York Times (August 12, Al, C20).
Cutter, S. L 1995. Race, class, and environmental justice. Progress/n Human Geography 19(1): 111-22.
Cutter, S. L., and W. D. Solecki. 1996. Setting environmental justice in space and place: Acute and chronic airborne toxic releases in the Southeastern United States. Urban Geography 17(5): 380-99. Cutter, S. L, D. Holm, and L Clark 1996. The role of geographic scale in monitoring environmental justice. Risk Analysis 16(4): 517-26.
Davies, J. C. (ed.). 1996. Comparing environmental risks: Tools for setting government priorities. Washington, D.C.: Resources for the Future.
Glickman, T. S. 1994. Measuring environmental equity with geographical information systems. Renewable Resources Journal 12(3): 17-21.
Glickman, T. S., D. Golding, and R. Hersh. 1995. GIS-based environmental equity analysis: A case study of TRI facilities in the Pittsburgh area. In: G. E. G. Beroggi and W. A: Wallace (eds.), Computer supported risk management. Dordrecht: Kluwer Academic. pp. 95-114.
Goldman, B. A. 1991. The truth about where you live. New York: Random House.
Harris, T., and D. Weiner. 1996. GIS and society: The social implications of how people, space, and environment are represented in GIS. Technical report 96-7, NCGIA.
Kunreuther, H., and P. Slovic (eds). 1996. Challenges in risk assessment and risk management. Ire Annals of the American Academy of Political and Social Science 545: 8-183.
Kuhn, W. 1992. Paradigms of GIS use. In: P. Bresnahan, E. Corwin, and D. Cowen (eds), Proceedings, 5th International Symposium on Spatial Data Handling. 1: 91-103.
Livingston, M. 1996. EPA: S. C. industry polluted less in '94. The State, June 27. B1, B7.
Mayo, D. G., and P. D. Hollander (eds). 1991. Acceptable evidence: Science and values in risk management. New York: Oxford University Press.
McMaster, P. B., H. Leitner, and E. Sheppard. 1997. Risk assessment in the Twin Cities using geographic information systems. Cartography and Geographic Information Systems
Minard, R. A. 1996. CRA and the states: History, politics, and results. In: J. C. Davies (ed.), Comparing environmental risks: Tools for setting government priorities. Washington, D.C.: Resources for the Future. pp. 23-62.
National Research Council (NRC). 1983. Risk assessment in the federal government: Managing the process. Washington, D.C.: National Academy Press.
NRC. 1994. Science and judgment in risk assessment. Washington, D.C.: National Academy Press.
NRC. 1996. Understanding risk: Informing decisions in a democratic society. Washington, D.C.: National Academy Press.
Nyerges, T., M. Robkin, and T. J. Moore, 1997. Geographic information systems for risk evaluation: Perspectives on applications concerning human and ecological impacts. Cartography and Geographic Information Systems 24(3): 123-144.
Scott, M., C. Menzel, and S. L Cutter. 1997a. Relative risk profiles: A methodology for assessing community risk. Columbia, South Carolina: Department of Geography, Hazards Research Lab Discussion Paper #9.
Scott, M., S. L Cutter, C. Menzel, M. Ji, and D. Wagner. 1997b. Spatial accuracy of the EPA's environmental hazards databases and their use in environmental equity analyses. Applied Geographic Studies 1(1): 45431.
Stockwell, J. R., J. w. Sorenson, J. W. Eckert Jr., and E. M. Carreras. 1993. The U.S. EPA geographic information system for mapping environmental releases of toxic chemical release inventory TRI) chemicals. Risk Analysis 13: 155-64.
Sui, D. Z. and J. IL Giardino. 1995. Applications of GIS in environmental equity analyses: A multi-scale and multi-zoning scheme study for the city of Houston, Texas, USA. Proceedings, GIS/LIS '95 vol. 2: 950-959. Bethesda, Md.: ACSM and ASPRS.
United Church of Christ (UCC) Commission for Racial Justice. 1987. Toxic wastes and race: A national report on the racial and socioeconomic characteristics of communities with hazardous waste sites. New York: United Church of Christ
USEPA (United States Environmental Protection Agency). 1987. Unfinished busing: A comparative assessment of environmental problems. Washington, D.C.: USEPA, Report Number EPA 230-2-87-025a.
Wernick, I. K. 1996. Editorial: Better risk information for communities. Risk Analysis 16(5): 601-3.
Wernick, I. K. 1997. Community risk profiles: Background. < http://phe.rockefeller.edu/comm_risk/>.Appendix
Weighted Average Toxicity Calculation Procedure
1. Create file of facility releases containing unique facility number, name of chemical released, CAS number of chemical released, amount released.
2. Create file of chemical toxicities (available from http://www.cla.sc.edu.pallas2.tcl.sc.edu/geo/hrl/sctrap/toxfaqs/ toxicity.html).
3. Join facility release file to toxicity file.
4. Create file of total releases per facility.
5. Relate file of total releases per facility to facility release file.
6. Calculate the proportional amount each chemical contributes to the total.
7. Calculate the weighted toxicity for each release.
8. Sum the proportional toxicities for each facility.
9. Combine the total amount per facility file and the weighted average toxicity per facility file. The result is a file with unique facility numbers and each facility's magnitude and toxicity of releases.
Note that this process takes place at the Arc: prompt and within Tables. These two programs will be differentiated by A: and T:, respectively.
T: DEFINE facfile FAC_ID,4,5,b CHEM_NAME, 30,30,c CAS_NO, 15,15,c AMT, 10,10,i T_AMT, 10,10,i <--These items P_AMT,4,6,f, 2<-- will be P_TOX,4,7,f, 3 <-- calculated later T: (Add facility information) T: DEFINE toxfile HEM_NAME, 30,30,c CAS_NO, 15,15,c TOXICITY,4,7,f, 3 T: (Add toxicity information) A: JOINITEM facfile toxfile facfile FAC_ID AMT A: FREQUENCY facfile amt_per_fac Freq Item: FAC_ID Sum Item: AMT T: RELATE ADD Relation name: temp 1 Table Identifier: amt_per_fac Database name: INFO INFO Item: FAC ID Relate Column: FAC ID Relate Type: LINEAR Relate Access: RO T: SELECT facfile T: CALCULATE P_AMT = AMT/temp1//AMT T: CALCULATE P TOX = TOXICITY * P_AMT A: FREQUENCY facfile wat_per_fac Freq Item: FAC ID Sum Item: P TOX A: JOINITEM amt_perfac wat_perfac amtwat_perfac FAC_ID FAC_ID
Michael S. Scott is a post-doctoral research geographer and Susan L. Cutter is Professor of Geography in the Department of Geography, University of South Carolina, Columbia, South Carolina 29208. Phone: (803) 777-5234; Fax: (803) 777-4972. E-mail: <firstname.lastname@example.org>.
Thomson Gale Document Number:A54090787