Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Environmental justice: geospatial impacts of hazardous materials spills
(USC Thesis Other)
Environmental justice: geospatial impacts of hazardous materials spills
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
1
ENVIRONMENTAL JUSTICE:
GEOSPATIAL IMPACTS OF HAZARDOUS MATERIALS SPILLS
by
Claudia E. Avendano
A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
GEOGRAPHY
May 2014
Copyright 2014 Claudia E. Avendano
ii
DEDICATION
To Lisa (Ph.D. Schweitzer), Jenny (Ph.D. Swift), Karin (Ph.D. Huebner), and Debbie (Ph.D.
Natoli) for their friendship and guidance, for leading the way by example, and for being living
proof that academic angels exist. To my family, my roots, my core, my dreams, and the
unconditional inspiration. To Stevie, my son and my true teacher. To my partner, for grabbing
my hand and encourage me to dive the world, exploring both blue kingdoms. And last, but not
least, my friend Jorge Reyes Rodriguez de la Gala y Gala for twenty and more years of friendly
intellectual discussions. If not for all, I could not find the strength and inspiration to accomplish
this work.
iii
TABLE OF CONTENTS
DEDICATION ............................................................................................................................. ii
LIST OF FIGURES ......................................................................................................................v
LIST OF TABLES ..................................................................................................................... vii
ABBREVIATIONS .................................................................................................................. viii
ABSTRACT ..................................................................................................................................x
CHAPTER 1: INTRODUCTION .................................................................................................1
Background of the Problem ..................................................................................................... 1
Approach to the Problem ......................................................................................................... 4
Purposes of the Study............................................................................................................... 7
Research Questions and Design ............................................................................................... 8
Significance............................................................................................................................ 11
CHAPTER 2: LITERATURE REVIEW ....................................................................................13
CHAPTER 3: METHODS ..........................................................................................................24
Frequency of Spills, and Cluster Analysis ..........................................................................43
Hotspots Analysis, Gettis-ord-gi* Analysis ........................................................................43
CHAPTER 4: RESULTS ............................................................................................................52
Methodological Innovations ...............................................................................................52
Trends, Number of Spills ....................................................................................................52
HAZMATS Spills Frequency .............................................................................................55
Results Research Question One: .........................................................................................58
California as a Case Study ..................................................................................................60
Results Research Question Two: ........................................................................................65
CHAPTER 5: CONCLUSIONS .................................................................................................74
BIBLIOGRAPHY .......................................................................................................................77
APPENDICES ............................................................................................................................82
Appendix I: Inventory of Federal Portals Containing Databases and other Information relevant
to HAZMATS Spills .............................................................................................................. 82
Appendix II: Legal Framework of Environmental Justice .................................................... 84
iv
Appendix III: Types of Units Commonly Used in EJ Analysis............................................. 90
Appendix IV:Definition of Key Terms and Operational Definitions .................................... 94
Appendix V: Table of Nation-wide Unique Locations with annual Frequency of Spills >10095
Appendix VI: Table Hotspots with Z score =>1.29 or 0.90 Percentile for California .......... 97
Table of Contents (Countinued)
v
LIST OF FIGURES
Figure 1: Spatial Distribution of HAZMATS Spills in the U.S., 1998-2009 ............................... 29
Figure 2: Dallas Forth Worth Metropolitan Area. Distribution of Hazardous Materials Spills,
Severe Incidents and Major Evacuations ...................................................................................... 30
Figure 3 Chicago Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 31
Figure 4: Boston Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 32
Figure 5: Atlanta Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 33
Figure 6: Washington Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 34
Figure 7: South Florida Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 35
Figure 8: San Francisco Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 36
Figure 9: Sacramento Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 37
Figure 10: New York Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 38
Figure 11: Los Angeles Metropolitan Area. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 39
Figure 12: Detroit Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 40
Figure 13: Houston Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations .................................................................................................. 41
Figure 14: Los Angeles CSA Metropolitan Area. Distribution of Hazardous Materials Spills,
Severe Incidents and Major Evacuations ...................................................................................... 42
Figure 15: Gettis-Ord-Gi* Equations from Arc Tool Box............................................................ 45
Figure 16: Hot Spots Analysis Model ........................................................................................... 48
vi
Figure 17: Hot Spot Analysis with Conceptualization ................................................................. 50
Figure 18: Historical Trends of HAZMAT Spills per Year. Influence of Legislation in Time
Series Interpretation ...................................................................................................................... 53
Figure 19: Total Number of Spills Incidents per Year ................................................................. 55
Figure 20: HAZMATS and Serious Incidents .............................................................................. 56
Figure 21: People Evacuated Due to HAZMATS Spills .............................................................. 57
Figure 22: Trends. Total spills, Serious Incidents, Total People Evacuated ................................ 57
Figure 23: Map Unique Identifiers (HAZMATS spills) per Location, National Level ................ 59
Figure 24: Hotspots Analysis, 0.90 Percentile .............................................................................. 60
Figure 25: California HAZMATS Spills, 1998-2009 ................................................................... 62
Figure 26: Hotspots Analysis for California ................................................................................. 63
Figure 27: Hotspots Areas in California ....................................................................................... 67
Figure 28: Spatial Distribution of MHA and Poverty Levels ....................................................... 73
vii
LIST OF TABLES
Table 1: Methodological Comparison among Main Environmental Justice Studies. ................... 20
Table 2: Serious Incidents and People Evacuated due to HAZMATS Spills ............................... 56
Table 3: Summary of Population within Major Hotspots Areas by Race and Ethnicity .............. 68
Table 4: Summary of Total Population by Race in California ..................................................... 68
viii
ABBREVIATIONS
Agency for Toxic Substances and Disease Registry (ATSDR)
Census Transportation Planning Package (CTPP)
Civil Rights Act of 1968 (Fair Housing Act)
Community Emergency Response Teams (CERTS)
Council of Environmental Quality (CEQ)
The Comprehensive Environmental Response, Compensation, and Liability Information System
(CERCLIS)
Emergency Planning and Community Right-to-Know Act (EPCRA)
Emergency Response and Notification System (ERNS)
Environmental Impact Study (EIS)
Environmental Impact Review Process (EIR)
Environmental Justice (EJ)
Environmental Justice Determinations (EJD)
Environmental Protection Agency (EPA)
Federal Comprehensive Environmental Response, Compensation and Liability Act (CERCLA,
or “superfund”)
Federal Hazardous and Solid Waste Amendments (HSWA)
Federal Insecticide, Fungicide and Rodenticide Act (FIFRA).
Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA)
General Accounting Office Report (GAO)
Geographic Information Systems (GIS)
Hazard Communications Standards (HAZCOM)
Hazardous Materials (HAZMATS)
Hazardous Substances and Emergency Events Surveillance (HSEES)
Hazardous Waste Treatment and Storage or Disposal Facilities (TSDFs)
HAZMATS Information System (HMIRS)
Hotspots Vicinity (HSV)
ix
Major Hotspots Areas (MHA)
Material Safety Data Sheets (MSDS)
Modifiable Area Unit Problem (MAUP)
National Cooperative Highway Research Program (NCHRP)
National Environmental Policy Act (NEPA)
Nearest Neighbor Distance Analysis (NNDA)
Nearest Neighbor Ratio (NNR)
Office of Solid Waste and Emergency Response (OSWER)
Percentage of Population by Race at the tract level (PPR)
Pipeline and Hazardous Materials Safety Administration (PHMSA)
Pipeline Inspection Protection Enforcement and Safety (PIPES)
Resource Conservation and Recovery Act (RCRA)
Resources Conservation and Recovery Act (RCRA)
Sector Facility Indexing Project (SFIP)
Significant New Use Rules (SNURs)
Superfund Amendments and Reauthorization Act (SARA)
Supervisory Control and Data Acquisitions (SCADA)
Toxic Substances Control Act (TSCA)
Toxics Release Inventory (TRI)
Traffic Analysis Zones (TAZ)
United Church of Christ Commission for Racial Justice (UCC)
US Environmental Protection Agency (USEPA)
x
ABSTRACT
This study encompasses Hazardous Materials Spills (HAZMATS) occurred in the United States
between 1998 and 2009, also concentrating on California as a case study. This work expands the
base of empirical knowledge and observations of previous studies dealing with the geographical
location of Hazardous Materials Handling and Storage Facilities (HMHSF)
Integrated into the HAZMATS spills distribution analysis, a frontline approach is
proposed to address the more common methodological issues and constraints faced by the
environmental justice discipline, as reported by previous authors. This new methodology is
purposely designed to comply with the current legal requirements of the Environmental
Protection Agency (EPA) for Environmental Justice Determinations (EJD), so can be expanded
from theoretical to the applied realm. Two major methodological problems were identified
during the literature review: a) The precise determination of the best geographical extent (scale)
to define the area of study and the appropriate unit of analysis to capture the environmental
problem at hand, and b) the introduction of artificial boundaries caused by the use of population
census data aggregated at the tract, block, or ZIP code level. The author proposes, creates, and
tests an innovative Geospatial Tool, based in the study of “Hotspots” -or outliers- and the use of
a variable radius in order to address the “Modifiable Area Unit” problem and the introduction of
artificial boundaries by the census data offering a solution for the two more common
methodological issues that have been encountered by previous authors.
This thesis also explores distribution patterns and spatial relationships between the more
significant HAZMATS spills locations (“Hotspots”) and highlights communities with high levels
of poverty and social vulnerability where questions concerning environmental justice need to be
xi
urgently addressed. Future work is proposed to assess the presence of “Hotspots” as predictors
of future severe spills and catastrophic events.
1
CHAPTER 1: INTRODUCTION
This chapter provides a historical perspective on the effects that hazardous materials spills have
had on populations residing in close proximity of Hazardous Materials (HAZMATS) facilities at
which spills have occurred. This section summarizes the main findings of this research and
compares them with the historical trends, focusing not only on the identification, and
geographical distribution of the HAZMATS spills, but also on the socioeconomic characteristics
of the adjacent populations which are permanently exposed to these toxic releases. These
findings raise serious questions regarding environmental justice and social equity.
This chapter also introduces the theoretical framework, main definitions, and limiting
assumptions of this study and states the purpose and nature of this work. This work frames the
problem of fair handling and distributing of HAZMATS as an object of scientific study,
highlights the significance of this problem, and postulates three critical research questions. Each
research question is accompanied by a null hypothesis and a brief description of the
methodological approach used to address each question.
Background of the Problem
This study compiles HAZMATS spills occurring in the continental United States (U.S)
between 1998 and 2009 (Appendix I). The HAZMATS spills located outside of the continental
United States are beyond the scope of this project and have therefore been removed from the
dataset compiled for this study. The information utilized in this study was collected from
multiple government databases and reports as Listed in Appendix I. By law, these databases must
document HAZMATS spills wherever they occur. Spills can occur at the HAZMATS handling
facility, at any location along the transportation route, or during transportation mode transitions.
However, this thesis focuses on the distribution of HAZMATS spills in the absence of any other
2
geographical or transportation variable in order to reveal how the distribution of HAZMATS
spills has changed during the past ten years, nationwide. The author also provides a brief analysis
of the time series patterns in the Results section of this thesis that explore whether or not changes
in legislation, technology, type of transportation or type of facility affected the HAZMATS spills
dataset.
In the context of this thesis , places where HAZMATS spills have occurred with
frequencies of statistically high significance are called “Hotspots” and the term "significant
means that the incidence of the data in each location may be attributed to a cause (as opposed to
chance) with a high percentage of certainty. The body of existing literature demonstrates that
hotspots have traditionally been located in close proximity to the most vulnerable sectors of the
population. Their high levels of vulnerability and the lack of cultural political and economic
resilience among other aggravating factors exponentially increases the consequences associated
to toxic long-term exposure these populations suffer.
Furthermore, “Hotspots” tend to be clustered together in small geographical areas,
compounding the risk of continuous exposure from multiple sources and major catastrophic
incidents. Hotspots and hotspot clusters have been also geographically linked to disastrous spills
and severe evacuations (Feng, 2001). In this thesis a severe evacuation event is defined as any
evacuation causing more than 500 people to be displaced. If frequency of spills can be used as a
predictor of future severe spills then studying the frequency of small spills may allow severe or
catastrophic spills to be anticipated and perhaps prevented. This study includes an analysis of the
spatial correlation between places with high frequencies of spills and places that have
experienced serious evacuation events. Another critical question is whether or not there is a
strong correlation between hotspots and high levels of poverty. A high correlation would indicate
3
that impoverished residential communities are more heavily burdened by the environmental risks
associated with hazardous materials in comparison to the average American.
Previous environmental research has identified potential causes for the higher level of
risk imposed upon impoverished communities (Feng, 2001). One cause relates to the existence of
poor planning, zoning, and housing policies that allow or encourage the construction of
HAZMATS facilities in close proximity to vulnerable neighborhoods. These policies encourage
disadvantaged populations to establish communities around HAZMATS handling facilities, due
to cheaper rents and lower property values that are associated with the undesirable living and
environmental conditions caused by HAZMATS presence (Been 1990 in Feng, 2001).
Unfortunately, the presence of the hazards is not always evident at the time residents choose
housing, but it becomes evident over time if health problems related to continuous toxic
exposure emerge.
Existing literature also discusses the role of political, social, and economic challenges
that contribute to disadvantaged groups not having sufficient political power to control what
happens in their own neighborhoods (Schweitzer, 2008), (Cutter, 1997), (Feng, 2001). They are
usually disenfranchised minorities, who either choose not to vote due to cultural political
differences with their place or origin, or are simply not eligible to vote. Either way, as a result,
their interests may not be reflected in the political decision-making processes that affect their
neighborhoods. These populations are therefore known as “silent populations” with no political
voice in decisions over what happens in their own community. Economic disadvantage also
reinforces the lack of political will or political power regarding this problem; often the residents
of impoverished neighborhoods have to prioritize jobs over local environmental quality because
HAZMATS handlers and shippers can offer revenue and jobs for the community (Been, 1990).
4
The social and economic disadvantages continue to echo throughout the entire “cradle to grave”
cycle of hazardous materials, compounding the risk over the surrounding populations. (Anderton,
1994)
For example, Schweitzer (2008) points to the fact that during emergency evacuation
situations, the various social groups within each community could respond differently to
emergency plans because they have different compliance capability and because of their own
cultural risk perception. Also, disadvantaged social groups may not know how to effectively
communicate with authorities during official attempts at pre- or post-disaster management;
adding an extra level of complexity to a disaster-related situation, as these groups do not behave
as authorities expect, but react to the threat incorporating their own cultural responses. Poor and
minority communities have historically been identified as key locations where issues of
environmental justice need to be addressed. These communities are places where planning,
preparedness, and mitigation efforts need to be not only prioritized, but customized by
incorporating genuine community responses to achieve a realistic level of performance during
catastrophic events.
It is the opinion of the author that planning decisions and HAZMATS policies need to be
based on accurate, rigorous, and equitable science that gives an informed voice to these often
silent populations. Thus this study seeks to provide a deeper understanding of these phenomena
as well as insight into the unfair exposure of socially vulnerable populations discussing these
populations’ unique needs with regard to emergency and evacuation planning.
Approach to the Problem
Beginning in 1970, several U.S. studies have made serious claims of environmental
inequity regarding the spatial distribution of toxic and hazardous waste facilities among
5
neighborhoods of different social, economic, or racial status. The first publications alleging the
inequitable distribution of environmental hazards nationwide were released by the Council of
Environmental Quality (CEQ), (CEQ, 1971) (CEQ, 1997). Ever since, research on
Environmental Justice (EJ) theory and its applications has significantly advanced (Feng, 2001).
Early concerns about the inequitable distribution of HAZMATS were manifested also by other
government and private entities. For example, the General Accounting Office Report (GAO)
(GAO, 1983) found that three out of four communities hosting toxic and hazardous facilities
were located in close proximity to African American populations, and that at least 26 percent of
community residents had incomes below the poverty level. In 1987 the United Church of Christ
Commission for Racial Justice (UCC) (UCC, 1987) also concluded that race was the leading
indicator associated with toxic facility location. Research in the 1980s focused mainly on the
relationship between the siting of toxic and hazardous facilities and the socioeconomic variables
of the populations surrounding these areas (UCC, 1987). It was largely because of these studies
that the topics of environmental justice, environmental equity, and environmental racism
developed into formal areas of inquiry both for scientists and government agencies.
Recently, the UCC conducted a follow-up of their 1987 study (CEQ 1997). Using data
from the 2000 census, the UCC compared and contrasted the conditions from the 1980s with the
conditions reflected in the 2000 census. The UCC demonstrated that after almost 30 years, racial,
economic and ethnic disparities continue to exist. Additionally other studies have found that
transport spills and on-site releases are also geographically concentrated disproportionately in the
most socially vulnerable communities (Cutter, 1997). From the perspective of environmental
justice, a problem of fairness exists when one sector of the population is subjected to a
disproportionate share of environmental burdens. This perspective asserts (based in historic
6
observations) that people of color and people with low income are bearing more than their fair
share of HAZMATS risk exposure (Bullard 1990, and 1994). For many, the excess burdens
imposed on these populations could be constituted as environmental racism and therefore
considered as a violation of federal principles and civil rights.
In recent years, the definitions of “protected populations” and “socially vulnerable
communities” have been expanded to include those defined by age, disability, gender, religion,
class, race, income level, English proficiency level, and national origin (Appendix II). This
expansion was largely the result of Executive Order 12898, which was issued by President Bill
Clinton in 1994. (For a detailed explanation of the main laws and executive orders of the Legal
Framework for Environmental Justice, please refer to Appendix II). The EPA has established a
protocol to prevent environmental injustice in any given community through a methodology
known as Environmental Justice Determination (EJD) (EPA 1995 in Feng 2001). This protocol
is designed to identify locations where unfair environmental burdens occur, especially if they
occur close to vulnerable populations as defined by Executive orders 12898, Executive Order
13166, and Executive Order 13045. Best practices in EJD have been proposed by the National
Cooperative Highway Research Program (NCHRP), which has released an informative report on
effective methods for Environmental Justice Assessments (NCHRP, 2004). The report identifies
three crucial steps to developing an accurate EJD assessment:
• Identify the affected population;
• Estimate the nature and extend of the effects; and
• Assess whether the effects are equitable.
Incorporating these recommendations and following current best practices, this study
proposes a new and innovative GIS methodology designed and built to comply with the EPA’s
7
legal and methodological requirements for improved EJD. This methodology first finds the
geographic distribution of previous HAZMATS spills, and then describes the populations most
affected by these locations. Subsequently, the hotspots have been identified over a fourteen year
time-series covering the continental U.S., and, using California as case study, the populations in
close proximity to the hotspots have been categorized by race at the census tract level.
The nature and extent of the effects surrounding each spill will always depend on the
particular discharge, the chemicals involved, the volume of these chemicals, and the special
circumstances that resulted in the discharge. The methodology in this study does not intend to
evaluate the nature and extent of each specific spill event; it is designed to facilitate
interpretation of the broader spatial patterns related to these accidents. The overarching objective
is to develop an inventory of locations where the more serious hotspots occur, and of places
where the hotspots affect vulnerable populations disproportionally. The output of this study
should provide a clear layout of places with the worst EJ issues that need to be prioritized, where
detailed studies about specific discharge events, chemical types, and health effects are needed the
most.
Purposes of the Study
• To expand upon the base of empirical knowledge gained from previous studies of
HAZMATS spills. To use spatial analysis techniques as a means of exploring the relationship
between the geographic distribution of HAZMATS hotspots and their proximity to
communities with high levels of poverty and social vulnerability;
• To address some of the more common methodological issues in Environmental Justice
studies, (such as the Modifiable Area Unit Problem, or MAUP) faced by previous authors
within the discipline. To build upon the current Environmental Justice and social equity
8
discourse ongoing since the early 1970s with regards to HAZMATS handling and
distribution;
• To provide a mechanism to improve current professional practices for Environmental Justice
Determinations (EJD) through the application of the proposed methodology.
Research Questions and Design
Research Question One: What is the spatial distribution of the observed HAZMATS spills?
H1: HAZMATS spills cluster together to create “Hotspots” of toxic and hazardous
incidents during transport.
As part of this study a master database containing data for all accidental chemical
releases between 1998 and 2009 was compiled from reports of accidental releases, which are
made available by federal, state, and local agencies. Particularly, the data for this study were
obtained from four datasets (Appendix I):
• The Emergency Response and Notification System (ERNS);
http://www.epa.gov/records/policy/schedule/sched/060.htm
• The HAZMATS Information System (HMIRS);
http://www.dlis.dla.mil/hmirs/
• The Hazardous Substances and Emergency Events Surveillance (HSEES);
http://www.atsdr.cdc.gov/hs/hsees/
• The Census Bureau (Census Bureau TIGER layers), for demographic information
relevant to the area of study;
http://www.census.gov/geo/maps-data/data/tiger.html
9
First, this master database of HAZMATS spills (points) was geocoded so that it could be readily
analyzed using a Geographic Information System (GIS). Each spill was treated as a separate
event and mapped using ESRI’s ArcMap utilizing the X, Y latitude and longitude coordinates of
each reported spill geocoded in the previous step. This point data was plotted for the entire U.S.;
any data that fell outside of the continental U.S. was discarded, even if the spill originated in the
U.S.
Second, a unique analytical model was built in ArcMap to reveal the more apparent spill
distribution patterns. This model was used to perform a statistical analysis of the HAZMAT
spills in order to draw conclusions about the statistical significance of the spatial variation and
spill distribution patterns. The frequency of spills per unique identifier was calculated in order to
assess statistical significance. In this study a unique identifier was defined as a location at which
one or more events of HAZMAT spill have occurred within a radius of 100 feet, to identify all
discharges happening within the same HAZMAT facility independently of their exact location
within the facility (parking lot, storage, transfer unit, process area, etc.).
The third step was then to evaluate the frequency of HAZMAT spill events within each
given area to determine the area’s statistical significance. All spills within an area of 0.5 mile
radius (centered at each unique identifier) were evaluated. Subsequently, a hotspots analysis was
run to determine the local mean frequency for each area and to determine if this frequency was
significantly higher than the global mean frequency of events. These statistically significant
areas are named Hotspots. For the purpose of this thesis, the term “significant” is used to
designate the areas with spill frequencies equal to or greater than the 90th percentile. The
geospatial methodology and geo-statistical principles used for the identification of hotspots are
described in detail in the Chapter 3: Methodology.
10
Fourth, the analytical model was compiled into a new ArcMap Geospatial Tool for the
convenience of model users. Innovative flexibility was integrated into the model to allow the
units of measurement and the total area of analysis to be varied by the model user, which is
intended to help address concerns related to the MAUP. This ArcMap tool format is a useful
package for sharing the model with other researchers and disseminating this work in the
Geospatial community.
Research Question Two: How are HAZMATS spills distributed among social groups?
H2: HAZMATS spills occur at a disproportionately higher level in poor neighborhoods
and in neighborhoods of color. (Poor neighborhoods and neighborhoods of color as defined by
Executive orders 12898, Executive Order 13166 and Executive Order 13045).
The literature suggests that poverty is perhaps the most important factor contributing to
environmental injustice, even though poverty is only one of the characteristics that contribute to
political and social vulnerability (Been, 1990). Once the first research question was addressed
and hotspots were identified, the next step is to identify secondary spatial patterns among
hotspots to determine if they are randomly distributed throughout the study area, or if they
appear in specific communities. To achieve this, the hotspot map was overlaid with a Census
Bureau TIGER layer called “Places” which identifies boundaries of major cities and other
incorporated places (http://www.census.gov/geo/maps-data/data/tiger.html). The result of this
overlay was then overlaid with an additional layer called “Small Metropolitan Areas”
(http://www.census.gov/geo/maps-data/data/tiger.html) so the hotspots could be geographically
referenced to specific cities and metropolitan communities. Next, a 0.5 mile buffer was drawn
around each hotspot representing the surrounding area most likely to be negatively impacted by
these hotspots. Lastly, the resulting hotspot area is overlaid on the census tracts layer.
11
The overlay analysis is intended to identify census tracts that contain hotspots. It is
anticipated that the intersecting distribution of hotspots and disadvantaged populations will
highlight the need for increased planning and risk management for better handling of hazardous
materials spills within those identified areas. The socioeconomic and ethnic characteristics of
these census tracts should then be compared to the same characteristics of hotspot-free census
tracts. Conclusions are then drawn by comparing the percentage of vulnerable populations
directly affected by hotspots with the percentage of vulnerable populations not directly affected
by hotspots. Ideally percentages of the affected population should be equivalent for all income
ranges and ethnicities, which would reflect an even level of risk parity; however this result is
unlikely according to historical precedents.
Significance
This thesis is significant to both academics and practitioners who are in search of a more
ethical, safe, and equitable HAZMATS location sitting. The findings of this study corroborate
those of several previous studies raising environmental justice issues dealing with the
distribution, allocation, and management of HAZMATS exposure across different sectors of the
population (UCC) (UCC, 1987) , (Bullard 1990, and 1994) ,(CEQ 1997). The results of this thesis expand
upon existing knowledge and provide practical insights as to how some of the more common
methodological issues of sample resolution (such as the MAUP and introduction of artificial
boundaries) can be avoided. The models proposed herein are newly designed and aim to improve
on a few of the limitations of the current methodological approaches, thus improving the
accuracy and resolution of current models used for EJD. The models were carefully designed to
comply with the current EPA legal framework for EJD so that this research can be directly
12
applied by practitioners and authorities to locate HAZMAT facilities toward a nondiscriminatory
distribution of environmental burdens.
13
CHAPTER 2: LITERATURE REVIEW
This chapter compares the methodologies and chosen area unit for the four most significant
studies pertaining to the relationship between hazardous materials and EJ. Included are the UCC
study (1983), the UMASS Study (1980), the Been Study (1990), and the GAO Study (1995). All
four of these studies are longitudinal studies (10 years or more) that were conducted at the
national level.
Concerns pertaining to the siting of toxic and hazardous facilities were first addressed in
the 1970s in several studies, including a report released by the Council of Environmental Quality
in 1971 (CEQ 1971). These studies piqued the interest of minority communities, whose members
felt that they were exposed to a disproportionately high level of risk. Minority leaders then
organized their communities and pushed political figures to expand the purview of civil rights to
include issues related to environmental justice (Feng, 2001).
However, it was not until the 1980s that literature began to expand by focusing more
closely on the relationship between HAZMATS geography and the socioeconomic
characteristics of populations living in close proximity to HAZMATS areas. Three sets of reports
helped spearhead the “environmental racism” movement and functioned as the cornerstones of
the environmental justice field of inquiry. These trailblazing reports focused on the effects of
toxic releases and the socioeconomic and racial variables of the population at major risk
according to their geographical vicinity to HAZMATS facilities: (1) The General Accounting
Office (GAO 1983); (2) The United Church of Christ Commission for Racial Justice (UCC
1987); (3) Bullard (1990), Burke (1983), and (4)Anderton, et.al (1994).
The General Accounting Office Report (GAO 1983) was the first major report to study
the correlation between hazardous waste landfills and the demographic characteristics of nearby
14
communities. The report discovered that three out of four case study communities had
populations that were majority African American, and at least 26 percent of the population in
each of these areas had incomes below the poverty level. This report was supplemented by a
comprehensive study at the national level, which was released by the United Church of Christ
Commission for Racial Justice (UCC 1987). The UCC’s study concluded that race was the
strongest indicator associated with the location of toxic facilities. Using this information, the
study further argued that race could be used as a predictor of hazardous facility location. The
UCC study compared the demographics of ZIP codes that featured no Hazardous Waste
Treatment and Storage or Disposal Facilities (TSDFs) versus the demographic information from
ZIP codes containing TSDFs. ZIP codes that did not contain TSDFs featured minority
populations that averaged 12.3%, while ZIP codes containing TSDFs featured minority
populations had nearly double that figure. The study also found that ZIP codes with one or more
TSDFs, or ZIP codes containing one of the five largest landfills, had the highest percentage of
minority residents, 37.6%.
As time passed, the three foundational studies that gave rise to the environmental racism
movement were criticized for their narrow perspective on race. Other EJ proponents had begun
to focus on other demographic characteristics, such as poverty and social class (references). New
data became available when Congress re-authorized the Superfund law in 1986 and added the
“Community Right to Know Act,” which mandated industries to issue annual reports regarding
the release of over 300 toxic substances. This information is contained in the Toxics Release
Inventory (TRI). Using this new data, researchers soon began to expand the study of populations
living in close proximity to environmental hazard sites. Bryant and Mohai (1992) presented a
review of 15 studies where the distribution of pollution was found to be inequitable by income,
15
and only one study where the distribution of pollution was found to be inequitable by race. These
developments were a main focus at the Conference on Race and Incidence of Environmental
Hazards, which was held at the University of Michigan in 1990 (Feng, 2001). This conference
provided a forum where scholars, researchers, and activists came together to share their findings
with federal and state officials as well as to discuss strategies for change (Bryant and Mohai
1992).
EJ research in the 1990s also became more localized and specific (Feng, 2001)Topics
addressed a wide variety of issues including youth’s exposure to lead, workers’ exposure to
pesticides, toxic substances and heavy metals in food sources, and a variety of other issues.
However, in spite of such a wide variety of topics, there was a general shift towards more
quantitative geographic analyses (Feng 2001). Bullard (1996) examined two of the three
cornerstone studies in environmental discrimination under a newer, broader approach. He took
the highlights about social and ethnic disparities found by the GAO (1983) and the UCC (1987)
and triangulated them with his own study involving five different communities in Texas. This
approach led to the first lawsuit in the United States that alleged environmental discrimination
under the 1964 Civil Rights Act, and encouraged other victims of environmental injustice to step
forward (Feng 2001).Years later, the UCC also revisited its original study (UCC 1987) by
incorporating data from the 2000 census. The updated study demonstrated that racial, economic,
and ethnic disparities continue to prevail after more than three decades (Feng 2001). The GAO,
UCC, and Bullard studies are of special importance because they helped direct attention to the
issue of environmental injustice and also helped justify grassroots organizations’ efforts to resist
the further placement of toxic and hazardous facilities in their communities (Feng 2001). These
groups’ allegations of environmental inequity were criticized for including anecdotal information
16
in their analyses and for applying simplistic analytical methods to demonstrate disparities. For
example, early studies used elementary techniques such as cross tabulations and basic correlation
analysis, assuming a linear causality that was hard to defend (Feng, 2001).
In the 1990s, many new studies emerged in the field of environmental justice. The
academic community gradually reduced the scope of these studies towards small geographic
areas and placed an emphasis on regionally-based studies, where conclusions could be more
defensible. The shift to regional and urban-scale studies altered the research from proving or
disproving environmental racism as a concept to a more comprehensive approach that explored
the reasons behind such racism. EJ research in the 1990s can be divided into three main areas of
study: environmental racism, hazardous waste facilities site placement, and HAZMATS spills.
Over time, the geographic and spatial methods used to study environmental racism have
become progressively more sophisticated. Authors such as Talih and Fricker (2002) used a
different two–part analytical framework to further understand the interactions among the
variables that cause the observed spatial correlation between hazards and specific population
groups: The first module was outcome-based and identified the potential associations among
race, socioeconomic status, and proximity to hazards; and the second module was process-based
and determined how this association may have occurred. These methodological advances gave
rise to new sets of questions and new areas of inquiry. Questions include: are economically
disadvantaged populations moving toward toxic and hazardous sites, or are the facilities moving
to these communities because the surrounding areas have less cultural, political, and economic
power to resist their presence? Are the toxic and hazardous industries attracting people from a
specific race or economic status as employees? Are vulnerable populations the only groups that
remain after groups with more political and economic power decide to leave an area? Are the
17
locations of hazardous facilities detrimental to the surrounding real estate markets? Is this
phenomenon therefore responsible for devaluating medium home values and rental prices in the
area, thereby attracting even more people from economically disadvantaged backgrounds to live
there? Could these groups oppose any economic activity, toxic or not toxic, living in such a stage
of poverty? And why aren’t Superfund sites directly compensating these populations? This study
attempts to address many of these highly relevant questions. The Talih and Fricker’s (2002)
study became significant because it changed the EJ paradigm from a model based solely on
racial discrimination to a model based on a multidisciplinary perspective. This paradigm shift is
the genesis to the current field of inquiry and also led to the introduction of GIS for the study of
events throughout time and place.
Following this state-of-the-art framework, authors began to draw more complex
conclusions and attain a more in-depth understanding of environmental justice and
environmental hazards. In her doctoral thesis, Schweitzer (2004) summarized 10 years of
empirical literature on environmental justice and demonstrated that there is ample evidence in
support of Talih and Fricker’s paths of causation. This evidence includes the following:
• Low factor prices attract both industry and low-cost housing to the same places (Boone
1999);
• Disadvantaged groups have less power to influence zoning and permitting decisions (Greer
1995);
• Industry and government intentionally target minority communities because they are less
able to resist (Wright 1997 in Schweitzer 2004);
• Poor communities, desperate for jobs and tax revenue, take any type of economic
development that comes their way even if it means heavy industry (Schnaiberg 2001);
18
• Low income and minority residents, who face discrimination in housing markets, move next
to industrial areas because that is the best (or only) housing that they can get in competitive
and discriminatory housing markets (Boone 1999); and
• White residents moved to suburban locations, which concentrates minorities and low-income
residents near heavy industrial sites at inner cities.”
These findings are echoed in a growing body of regionally-based studies. Boer, et.al (1997)
studied the location of hazardous waste TSDFs in Los Angeles, CA, and concluded that the
communities most affected by TSDF are working classes and communities of color located near
industrial areas. Mohai and Bryant (1992) studied EJ issues in three counties in and around
Detroit and found strong correlations between race, economic status, and the distribution of
hazardous waste facilities. Perlin et al (1995), developed a study of toxic air emissions at the
county level and showed disparities by race, ethnicity, and – to a lesser extent – by income.
Pollock and Vittas (1995) studied the distribution of Toxic Release Inventory (TRI) facilities in
Florida by measuring seven variables at the block group level, thus proposing a logarithmic
function of proximity as the best measure of exposure. Brooks and Sethi (1997) examined the
distribution and density of TRI facilities by using ZIP codes as the unit of analysis for
population. They also developed an index that incorporated TRI toxicity as a variable dependent
on the distance from emissions and found racial biases were more pronounced than income
biases.
Later, regionally-based EJ studies also suffered from methodological criticisms. Authors
worked without formal conventions and were using different geographic scales and different
areal units for each study. This made it almost impossible to compare results among studies.
Also, many of these regionally-based studies lacked strong mathematical and statistical methods
19
to support the authors’ claims. With a few exceptions, studies were hampered by serious
limitations with respect to both the availability of historical data and the geospatial scales of
analysis. These hurdles seriously impaired the previous studies’ ability to report consistent
conclusions, and indicate that a more rigorous set of spatial methods was necessary.
During the evolution of the EJ movement, science experienced several notable paradigm
shifts from focusing on the socio-economic and environmental circumstances of the affected
populations, to the response of said populations to those challenges, the participation of civil
rights activists, and even to civil rights. Although most of these changes were consequence of
new knowledge, they have nonetheless generated controversy within the EJ field, as proponents
of different paradigms have antagonized and contradicted each either’s findings. The EJ
movement started with a single approach rooted in the discrimination/racism model, where three
major longitudinal studies complemented and, at times, contradicted one another. Over time, the
movement morphed into a multi-theoretical mode of inquiry with a myriad of improvements and
challenges related to the development of new methodological approaches. Problems of scale and
the specific circumstances surrounding each case have also resulted in major setbacks for the EJ
discipline, thereby making generalization and the verification of findings between studies very
difficult. This in turn has resulted in a lack of consensus and, at times, even animosity among
various proponents of the EJ movement.
Table 1, adapted from Feng Liu (2001), graphically represents this problem by using the
three EJ pillar studies and highlighting their differences in scope, methodology, and size of the
chosen area unit.
20
Table 1: Methodological Comparison among Main Environmental Justice Studies.
Study: UCC UMASS Been
Year 1980 1980 1990
TSDF’s 415 446 608
Unit of Analysis 5-digits ZIP Code Census Tracts Census Tracts
Number of Host 369 408 600
Definition of Race
Minority represented
by Hispanic and non-
Hispanic nonwhite.
Blacks, African
Americans &
Hispanic
Minority defined as
other than white and
all Hispanics
Income: Mean household
income
% of families below
the poverty line
Median Family
Income
Control Variables Mean value of owner
occupied homes.
Mean value of
housing stock
Median house value
Pounds of
HAZMATS waste per
person
% males employed in
labor force
% workers in
manufacturing
industry
TSDF’s per 1000
person’s
Mean population
density
Statistical Methods Discriminant
Analysis
T test T test
Difference of means Wilcox rank sum test
Difference of means
non parametric
Logistic regression Logistic regression
Inequity by Race Yes No Yes Hispanic,
No African American
Inequity by Income Yes Yes for bivariate
No for multivariate
Yes for bivariate
No for multivariate
Race more
Significant than
Income
Yes No No for bivariate
Yes for multivariate
Table 1 illustrates the lack of consistency – both in methodology and in the chosen unit of
analysis – that has become a major obstacle in drawing, defending and generalizing conclusions
across EJ studies. To some extent, these inconsistencies have made it difficult to replicate the
scientific methodology involved – which is a direct violation of the scientific method. These
issues are serious and therefore require further investigation. This section first addresses the
21
problems created by the use of different area units, and then proposes a model to address these
methodological concerns.
The UCC study compared the populations surrounding the sites of more than 410 TSDFs,
using ZIP codes as the preferred unit of analysis. UCC concluded that population groups
contained in ZIP codes with TSDFs (mainly minorities and economically disadvantaged groups)
were disproportionately burdened by HAZMATS. However, the UMASS study, which used
census tracts as the unit of analysis, produced markedly different findings. The discrepancy
between these two studies started a battle in the EJ field. Some authors, such as Anderton
(Anderton 1994), argue that the main differences between study results can be attributed to the
application of different area units. Anderson argues that ZIP codes are not the correct unit of
analysis. Authors including Mohai (Mohai 1995) express strong concerns about both ZIP codes
and census tracts, arguing that tract units were perhaps too small for the analysis and ZIP codes
were perhaps too large. Mohai conducted his own analysis using a circle with a fixed radius of
2.5 miles, and produced results similar to those obtained in the UCC study.
To examine the adequacy of the unit of analysis, Gilman, Guild, and Hersh (GGH 1995)
compared the results of five EJ studies on air toxicity that used five different units of analysis:
block groups, census tracts, municipalities, and circular buffers of 0.5 and 1.0 mile radii. The
GGH study found that the area units were not always significant. For example, the authors found
that the relationship between poverty level and air toxin exposure did not change by adjusting the
unit of analysis, while the relationship between race and air toxin exposure changed significantly
by adjusting the unit of analysis.
Cutter, Holm, and Clark (1996) developed a similar study that used block groups, census
tracts, and counties as units of analysis to assess three different sets of data: national priority list
22
sites (NPL), Toxics Release Inventory (TRI), and TSDFs. This study found that only a small
disparity exists in the relationship between HAZMATS facilities and income level when using
census blocks and census tracts as units of analysis. Furthermore, this analysis found that the
disparity between HAZMATS facilities and income level seems to be lost when counties are
used as the primary unit of analysis (CHC 1996). Wrigley (1996) found that as the unit of
analysis varies and different scales of resolution are applied, differences in statistical findings
emerge. Wrigley identifies this phenomenon as the “Modifiable Area Unit Problem” (MAUP).
Lastly, Sur and Giardiano (1995) explain MAUP even further and divide the problem into
its two fundamental elements: scale and zoning. Sur and Giardiano tested the effects of both
scale and zoning in order to explain the TRI per capita. Their study found that as the level of
geographic resolution increases, the importance of minorities and population density both
increase; however, the importance of income decreases. Sur and Giardiano’s study also found
that with zoning (using a circle with a fixed radius), the area of analysis can also significantly
impact results and can potentially lead to different statistical findings. The most common units
used for Environmental Justice Analysis are discussed in Appendix III.
Using a variety of statistical models, authors such as Ringquist (1997) showed that
different statistical methods can be used to address specific environmental concerns and identify
control variables that function as good predictors of TRI facilities. These predictors included: (1)
manufacturing employment; (2) median housing age; (3) urbanization; (4) race and ethnicity; and
(5) income class.
More recently, the authors of EJ studies and the government agencies in charge of EJ
determinations have begun to incorporate toxicity data into the geographical associations among
HAZMATS and populations, which gives a more meaningful application to all of their EJ
23
evaluations (Feng, 2001). However, an assessment of the human health impacts of chemical
emissions across different sectors of the population is very challenging. Some methods have
already been developed to assess HAZMATS waste sites for certain chemicals and to deal with
particular health issues, but in general efforts to integrate HAZMATS data with demographic
data have been severely limited (NCHRP 2004).
The EPA’s Sector Facility Indexing Project (SFIP) is one of the more recent attempts to
fill this void (EPA 1995c). The SFIP also makes information available to the public through the
Internet in order to comply with the public’s “right to know.” Various methods have been
designed to incorporate accurate information regarding chemical emissions toxicity, persistency,
bioaccumulation, and exposure into this new index. However, the SFIP has not yet been fully
developed for all health effects and HAZMATS, and therefore does not reflect vulnerability of
the population. The effects of toxicity on populations represent a highly localized phenomenon
that depends on the particulars of the spill, the community where the event occurs, and the
characteristics of the population living within the affected community.
Important information was gleaned from the past studies reviewed in this chapter, with
particular attention paid to the problems created by the historical use of different area units.
Lessons learned from the previous studies were considered in this study in the creation of the
analytical model developed to analyze HAZMAT spill data, detailed in the next chapter.
24
CHAPTER 3: METHODS
The previous chapter debated some of the main methodological issues hampering the advance of
the Environmental Justice as a science, attacking the validity of various main studies because the
findings were not consistent across scales and population units, and sometimes, when the
scenarios where replicated but using different population units the findings where altered,
making it almost impossible to compare and defend results among studies and therefore
discrediting the findings (Feng, 2001). This chapter will focus on the evolution of the
methodologies developed to cure those discrepancies.
The use of Geographic Information Systems (GIS) becomes the next major
methodological leap into EJ determinations; GIS helps to establish the legitimacy of
Environmental Injustice allegations by allowing researchers to visualize, analyze, understand,
and graphically represent the spatial nature of HAZMATS spills in terms of their spatial
relationship to nearby populations. GIS technology made it much easier for researchers to
understand the distribution of environmental benefits, burdens, and economic spread of the
HAZMATS economy in relationship to “protected populations” (NCHRP, 2004).
This thesis successfully utilizes ArcMap GIS analysis tools to represent the spatial
behavior of HAZMATS spills. The information conveyed in this study is intended to provide a
better understanding of the distribution patterns of HAZMAT spills, and how the frequencies of
spills have evolved, migrated, and changed over time. Also, conducting an analysis of population
characteristics, light may be shed on the correlation between “protected populations” (or
minority populations) and HAZMATS facilities over time.
Buffer analysis (creating areas of specified width around one or more map features) is
another GIS technique utilized in this thesis to approximate the direct impact area of HAZMATs
25
spills or hotspots over the nearby population. Subsequently, as information about the
composition of HAZMATS spills and the geographical extent of their health effects become
available, a more detailed analysis could be performed over each buffered area. For example, in
cases where HAZMATS spills have involved several different chemical types, each chemical
type can be represented by a layer with a buffer width representing the extent of the health
effects of each different chemical contained in the spill. Subsequently, an overlay analysis of
these buffer areas can be used to estimate the compounded effects of each layer and the
environmental effects they represent.
Mapping clusters is another valuable GIS tool for analyzing compounded effects of
several hotspots present within a geographical region, in situations when is necessary to draw
conclusions or take action based on the combine effects of those HATZMATs spills within a
small geographical area. Areas where these clusters occur are places where environmental
conditions are exacerbated and the associated risks are increased in exponential proportions. The
location of a HAZMATS spills cluster is also important in seeking the causes of clustering and
providing policy recommendations to reduce clustering and therefore reducing exponential risk.
Relief funds can be distributed more efficiently when the funds are allocated to places with the
largest probability of a HAZMATS event occurrence and the largest concentrations of at-risk
populations.
Hotspot analysis and co-clustering methods can be combined to create local estimates of
the intensity of a particular phenomenon. This thesis defines the number of events per “unique
location” (the frequency of events) as an indicator of intensity. This study compares the “local
mean frequency” (attained averaging the frequency of events within the 0.5 miles contiguous
area) to the “global mean frequency” (calculated for the total events in the sample). This
26
comparison makes it possible to determine the statistical significance of a point and to determine
if that point is a “Hotspot”, a cold spot, or an outlier by a gettis-ord-gi* statistical analysis.
This thesis proposes a particular methodology using a customized version of the gettis-
ord-gi* technique readily available in the ArcMap Toolbox. This particular tool first looks at all
of the points showing high-frequency, and then determines which of these points are of statistical
significance. An analytical model has been created by combining two ArcGIS spatial tools,
namely Hotspots analysis and Buffering. The resulting model allows the user to specify different
radii for consecutive analysis in order to achieve consistency while running the analysis for
different scales or geographic regions. Running the models at different spatial scales and
comparing the proportionality of the results among runs will help to determine if the scale of
analysis is negatively affecting the analysis; if the results among runs are similar at different
scales, then the different scales of analysis are not significantly affecting the results. However, if
the results vary among runs, the use of different scales is in fact altering the analysis, therefore, it
would be necessary to identify which scale provides the best detail-resolution according to the
focus and scope of this particular analysis.
To allow the maximum level of resolution, the model has been provided with the capacity
to dissect and calculate only an arithmetical proportion of the total area of a tract, whenever the
totality of the tract is not fully contained within the user defined area. The combination of these
two capabilities (allowing the user to vary the length of the area to be studied several times while
maintaining the rest of the variables constant, and the ability to take only an arithmetical
proportion of a tract) work to filter interference by geographic scale of area of analysis as well as
the introduction of “artificial boundaries” when using uncut tracts for census population analysis.
27
In this way, our methodology addresses two of the main problems associated with past
Environmental Justice studies; the problem of scale and the Modifiable Area Unit.
To address issues of data quality, careful consideration was given to the most common
issues affecting this type of dataset. This is particularly important, since the quality and
consistency of data could also affect the validity of the studies. In cases like this, relying on
government databases, consistency is a strong concern. The availability and continuity of
information can be adversely affected over time due to several factors; the content and format of
the government datasets often change in response to changes in legislation or because of budget
constraints, the data collection often gets discontinued or suffers from significant variations in
the quality of the data, electronic format, or distribution policies causing for them to be
extremely challenging to compare across database versions.
For example, as a result of a change in legislation in 1990 the EPA added 25 chemicals to
its original list of hazardous materials, which had originally only included 18.Therefore,
although the production of HAZMATS materials remained relatively constant, the number of
spills reported and the number of facilities monitored multiplied significantly in one year
because more waste was considered hazardous than in previous years. The increase in the
number of spills reported obeyed mainly to changes in the legislation, rather than reflecting a real
increase in production of hazardous materials spills. Similar changes in legislation have occurred
in 1992 and 1995, according to the EPA (EPA 1995c).
Strong consideration was given also to identify user mistakes, as the process of spill
reporting is made entering the data manually into electronic forms and it is not uncommon to
find misspelled words and discrepancies regarding the names of the facilities or the chemical
components of the spill, as well as the format of the individual entries. Also, it is imperative to
28
give some consideration to the correct use of projections in setting up the GIS analysis, software
limitation and the units utilized in the analysis.
Software limitations are another common cause of error. During this analysis, there were
serious limitations and interface problems while using Microsoft Excel to analyze relatively large
datasets especially when interacting with ArcGIS software. Excel lacked the capacity to handle
files of more than 10,000 records and when this threshold was exceeded, truncation and file
corruption errors commonly occurred. It was necessary to create new database handling system
in SQL, a more complex but powerful alternative. A QA/QC analysis was run and observations
were carefully recorded as metadata to ensure the completeness and accuracy of the entire
dataset at each step of the process. The master database was also examined to identify extreme or
missing values, locate mistakes in original reporting, and to ensure consistency with regard to
geographic extent of datasets, avoiding incongruities on the analysis. Once the database was
functional, every spill was plotted as a single event generating a first basic visualization model to
reveal apparent distribution patterns, while exploring issues created by the Modifiable Area Unit
problem. The next step involved selecting instances of statistical significance within the general
distribution patterns of HAZMATS spills; this is accomplished by identifying hotspots and
performing a hotspot cluster analysis. For the correct interpretation of the result of this analysis,
a careful explanation of the key term and operational definitions used by the author to define
hotspots and statistical significance is provided in Appendix IV. Figure 1 displays the first map
visualization of the input data for this study. This map shows the georeferenced HAZMATS spill
locations, hotspots and evacuations compiled in this study’s master database. This figure
illustrates the basic spatial distribution of this hazardous spill data set in relation to places where
major evacuation events had occurred.
29
Figure 1: Spatial Distribution of HAZMATS Spills in the U.S., 1998-2009
Additional detailed map visualizations were generated for each of the nine greater
metropolitan areas as defined by the U.S. Census. The following maps presented in Figures x
through x portray the relationships between hotspots, HAZMATS spills, and major evacuations.
These maps allow more detailed examination of the individual areas most affected by
HAZMATS spills, as per the input dataset.
30
Figure 2: Dallas Forth Worth Metropolitan Area. Distribution of Hazardous Materials Spills, Severe
Incidents and Major Evacuations
31
Figure 3 Chicago Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
32
Figure 4: Boston Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
33
Figure 5: Atlanta Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
34
Figure 6: Washington Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
35
Figure 7: South Florida Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents
and Major Evacuations
36
Figure 8: San Francisco Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents
and Major Evacuations
37
Figure 9: Sacramento Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
38
Figure 10: New York Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
39
Figure 11: Los Ángeles Metropolitan Area. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
40
Figure 12: Detroit Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
41
Figure 13: Houston Metropolitan Areas. Distribution of Hazardous Materials Spills, Severe Incidents and
Major Evacuations
42
Figure 14: Los Ángeles CSA Metropolitan Area. Distribution of Hazardous Materials Spills, Severe Incidents
and Major Evacuations
43
Frequency of Spills, and Cluster Analysis
As a first step for the hotspots analysis, the number of spills per location was counted, the
statistical significance of the number of spills per each location was assessed, and the locations
showing a statistical significant number of discharges were collected as a separated data subset
and labeled as hotspots. To be declared as a hotspot, the location needed to reach the threshold
value of Z = 1.29 pre-established in this study to declare a HAZMATS spills location a
“Hotspot.” The value of Z = 1.29 was chosen because a Z=1.29 is equivalent to 0.90 percentile,
ensuring that at least 9 out of 10 (90%) of the locations will have less discharges than a point
labeled as a Hotspot confirming that the point is a hotspot with a 90 percent of certainty. Once
generated the newly compiled hotspots dataset was used to map and examine the spatial
distribution of general the HAZMATS spills versus the hotspot distribution.
Several hotspots can occur “clustering” within a small geographical area; the mapping of
clusters is particularly useful to identify spatial outliers, to assess the causes of clustering, and to
estimate the compounded effects of multiple polluters within the same area. Hotspot analysis and
co-clustering methods can help answer questions about compounding effects by creating local
estimates of HAZMATS spills intensity and comparing them with the average values attained for
the whole sample (continental U.S). This study equals the frequency of spills with a first measure
of intensity; however, a complete analysis should be performed incorporating other significant
variables such as the volume of release, the toxicity of the hazardous materials involved, and
other parameters that can determine the magnitude of the event with more precision.
Hotspots Analysis, Gettis-ord-gi* Analysis
In addition to finding the number of events per unique location (frequency of spills), it is
necessary to distinguish as hotspots the locations with significantly high frequencies of spills, not
44
only with respect to their neighbors but also with respect to the rest of the HAZMAT handlers in
the total sample (Continental U.S.). Included in the ESRI Arc Map 10 toolbox is a statistical
analysis tool called hotspot analysis, or Gettis-ord-gi,* which evaluates every unique location
and finds the ones with high frequency of spills and the ones with statistical significance. The
Gettis-ord-gi* tool gather and averages the frequency of spills at each unique location within an
area defined by the user and labels it “local mean’. This Tool finds the “local mean” value by
integrating the frequencies of all spills at each location within the user defined area, and
averaging them over the total number of unique locations within the area. This analytical tool
also generates a “global mean” by averaging the frequency of each unique location for the entire
sample (Continental U.S.) and calculates a “global mean frequency,” which is referred to as the
“global mean.” The Gettis-ord-gi* tool compares the “local mean” with the “global mean” to
determine the statistical significance of each “unique location” and its vicinity. This information
is used to determine whether the user defined area is a “Hotspot”, a cold spot, or an outlier.
To determine the statistical significance of these areas, it is necessary to calculate a
normalized value, or Z value, for each polluting area. The Z value is based on the number of
incidents per location and the degree of certainty that this distribution is attributed to a specific
cause, as opposed to chance. The null hypothesis for pattern analysis tools essentially states that
there is no spatial pattern among individual events, (basically denying the clustering of
individual spills as a pattern of statistical significance). However, the Z score helps reject this
null hypothesis and demonstrates that there is a significant cause behind the clustering of spills.
Lastly, another value, the P value represents the probability of having falsely rejected the null
hypothesis; the P value assumes that the distribution pattern occurred by chance.
45
This information and the formulas to perform this calculations are conveyed in the
following figure (Figure 15) showing the equations and parameters to calculate the Gettis-Ord-
Gi* statistics.
Figure 15: Gettis-Ord-Gi* Equations from Arc Tool Box
http://webhelp.esri.com/ARCGISDESKTOP/9.3/index.cfm?TopicName=Hot_Spot_Analysis_%28Getis-
Ord_Gi*%29_%28Spatial_Statistics%29
The width of area to analyze surrounding (or buffering) each “unique location” it can be
freely determined by the user in order to create different analysis scenarios providing the user
with the maximum flexibility and accuracy while delimiting the area to be studied. However, the
tool has also the capability to recommend an appropriated area of analysis by using the “Nearest
Neighbor” method. In principle, any spatial point with a high rate of spills can be considered a
hotspot; however, the objective of this analysis is to find “regions” with statistically significant
high spill rates. This distinction is made possible by looking at each location with a high
frequency of spills in the context of the adjacent locations. To be statistically significant, a
hotspot location needs to have a high Z value and it needs to be surrounded by other features
with high frequency of spills. Next, the summation of high-frequency-of-spills locations and
46
their neighbors is calculated and labeled “local sum”. The “local sum” then is compared
proportionally to the sum of all features “total sum”. If the local sum is found to be much
different than the total sum, and the difference is too large to be the result of chance, then a
hotspot of statistical significance has been identified using the Gettis-ord-gi* analysis.
The determination of the geographic boundaries for a hotspot analysis is also critical. The
Gettis-ord-gi* equation assigns weights to each polluting location with respect to its neighbors,
giving more weight to the closer neighbors and less weight to the points far away according to
the threshold distance defined by the user. There are several methods to assign this relative
weight, the “gradual decay” of value proportional to distance from the center of the cluster, the
delimitation of the values by the” nearest neighbors” ,or the “Delaunay triangulation”.
This study ran the model using a one half mile fixed-radius distance as the “user-defined”
area unit of analysis. The threshold distance is defined as 1/2 mile, because is one of the units
must frequently used by other authors, allowing for meta-analysis and comparison of the results
of this study with the results of other studies. Furthermore, this unit most closely represents the
equivalent of a census tract; has been agreed upon as a good representation of a community; is
considered to introduce fewer aggregation errors than other units such as ZIP codes; and has
proven to be appropriate to represent the effects of different environmental hazards, as found in
the literature review.
The ½ mile level of analysis was performed at the national level and for each of the main
nine metropolitan regions to find hotspots at these different geographic scales. To determine the
adequacy of the ½ mile distance as unit for the analysis for each scale incremental values of at
least two different radii, 0.5 and 1 mile were tested. By testing at least two different distances as
47
radius, it will be determined if the size of the radius significantly alters the results of the analysis
and if so, how the results are altered.
Two spatial models were created using ArcGIS Model-Builder to streamline multiple
analyses, using the hotspot analysis Gettis-ord-gi*. The models were customized to allow for
users to control one or more parameters at the time for each run without altering any of the other
variables associated with the analysis. In addition, these models automate the production of
graphics and maps. A first map is produced by showing the polluting locations and their
respective centers. A second map is then produced using a color scale assigned to each location
as a function of its Z value, with a graduated symbol range of values using different symbol
sizes. This symbology values represent the number of events grouped by location. The total
sequence of steps and the process followed by the models are illustrated on the following page,
in Figure 16.
Hotspot Analysis Model Simple
The first module is called Hotspot Analysis Simple, as shown in Figure 16. The main
components of this model are described as follows:
• Copy Features: Copies the original file to an internal file to work inside the model. This is
done to avoid an instance in which the modules override any original files in the process;
• Integrate: Clusters the events that are within the threshold distance as indicated by the user,
in this study set to 0.5 miles, and then it calculates the center of the cluster;
• Collect: Counts how many spills were accumulated in the cluster;
• Add Field and Calculate Field: Adds a field and assigns a unique identifier for each cluster,
and creates a shape file called Weighted Spills.
48
Figure 16: Hot Spots Analysis Model, created using ArcMap Model Builder.
49
Hotspot Analysis Model with Conceptualization
The input data for this is model is already contained in the file named Weighted Spills,
which was generated as an output of the first model at the Copy Features process. The Add Field
and the Calculate Field functions re-cluster the polluting locations and create a spatial weights
matrix. The spatial weights matrix is then used as input to perform a Gettis-ord-gi* Statistic
Analysis, which will locate hotspots by calculating the Z score and P value for each cluster of
polluting locations. This process is illustrated in Figure 17, which is located on the following
page. Finally, the Z Score Rendering Module creates a layer file to be displayed as a map. This
map identifies the hotspots, using both a color scale assigned to each location as a function of its
Z value and a graduated symbol scale using different symbol sizes to represent the number of
events grouped in each location.
50
Figure 17: Hot Spot Analysis with Conceptualization, created in ArcMap Model Builder.
51
Nearest Neighbor Distance Analysis of Hotspots
The Average Nearest Neighbor Distance tool is used to measure the distance between
each hotspot and its nearest neighbor for the whole dataset. This tool then adds and averages all
of these nearest neighbor distances among hotspots. If the average distance among hotspots on
the sample is less than the average distance for a hypothetical random distribution, then the
distribution of the hotspots being analyzed is considered to be clustered. On the other hand, if the
average distance of the analyzed sample is greater than a hypothetical random distribution, then
the hotspots are considered to be dispersed. This dispersion or convergence is expressed as an
index, representing the ratio of the observed distance among hotspots divided by the expected
distance from a randomly distributed sample. If the index is less than 1, then the pattern indicates
clustering; conversely, if the index is greater than 1, then the pattern indicates dispersion.
To demonstrate these findings with statistical confidence, it is necessary to calculate the
Z value for each Nearest Neighbor Distance. The Z score value is a measure of statistical
significance that indicates whether or not the null hypothesis should be rejected. In this case, the
null hypothesis states that the points are randomly distributed.
52
CHAPTER 4: RESULTS
The following section describes the results of the model run, and summarizes the
methodological innovations introduced by the application of the designed GeoTools; then
contextualize the results with trends and spills geospatial distribution within the United States,
and the last section describes the results attained while using California as case study.
Methodological Innovations
The research presented in this thesis closes two gaps in the existing body of research.
First, there is currently no national study of hazardous materials spills. This fourteen-year
longitudinal study assesses the impacts of spills nationwide, thereby helping to close this gap.
This study is focused in the spatial distribution of high frequency spills that occur on the same
location and their effects over adjacent populations. Second, this thesis proposes a methodology
to study the relationship between these spills and the vulnerable populations affected by them
using several scales but without confining the analysis to the Census tract level resolution. This
helps address some of the more common methodological issues encountered by previous authors.
One purpose of this study is to incorporate these methodological advances into the current EJ
research field and to current professional practices in order to encourage positive change in the
current patterns of environmental inequity.
Trends, Number of Spills
Cutter and Minge (1997) found that between 1971 and 1980, the number of spills
increased steadily, with a peak in 1978. This study finds a similar trend, where the number of
spills between 1998 and 2009 gradually increased, with a peak in 2006 as conveyed in Figure 18.
53
Figure 18: Historical Trends of HAZMAT Spills per Year. Influence of Legislation in Time Series
Interpretation
The peak number of discharges in 2006 can be associated with changes in legislation that
occurred in the wake of the Exxon Valdez oil spill, a major oil spill that occurred in Alaska. In
response to this disaster, the U.S. House Energy and Commerce Committee passed a new series
of legislation on September 27, 2006 entitled the Pipeline Inspection Protection Enforcement and
Safety (PIPES) Act. The PIPES Act required federal officials to regulate virtually all "low-
stress" oil lines. Prior to the passage of the legislation, only high pressure oil lines that run under
heavily populated areas were monitored. The Pipeline and Hazardous Materials Safety
Administration (PHMSA) issued an initiative to expand its supervision to rural, low-stress lines
in "unusually sensitive" areas that cross environmentally sensitive areas or contain endangered
species.
The PIPES Act broadened the scope of the systems-based approach to assess and manage
safety-related risks and increased the number of personnel supervising and enforcing
HAZMATS reporting nationwide. The additional initiatives associated with the Act included:
• Increasing enforcement activity, transparency, and data quality;
NUMBER OF SPILLS PER YEAR
0
5000
10000
15000
20000
25000
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
YEAR
NUMBER OF INCIDENTS.
NUMBER OF SPILLS
54
• Implementing an integrity management program for distribution pipelines; and
• Requiring a human factors management plan to reduce risks associated with operator fatigue
in pipeline control centers; and
• Implementing National Transportation Safety Board recommendations on the Supervisory
Control and Data Acquisitions (SCADA) systems in pipelines;
Since the establishment of the PIPES Act, PHMSA has doubled its enforcement, increased
transparency, improved data quality standards, and toughened proposed pipeline safety civil
penalties. The average number of penalties per case has more than tripled since 2006.
A causal relationship of social, political, and economic issues regarding the occurrence of
HAZMAT incidents and reports has been previously documented (Cutter, et al. 1997). Therefore,
it is probable that that the PIPES Act and the associated appropriations may explain the sudden
increase in discharge reports in the year 2006.
55
HAZMATS Spills Frequency
Relationship between Severe Spills and Total Spills: The numbers of both serious spill events
and the people evacuated due to severe spill events show a direct relationship to the number of
discharges. First, both datasets display a decadal descendent trend. Second, neither of the
datasets showed the 2006 peak that is present in the number of discharges data – a point that is
demonstrated in Figure 19 to 22, and Table 2 below.
There is no significant correlation between the number of serious spill events and the number of
people evacuated due to severe spill events.
Figure 19: Total Number of Spills Incidents per Year.
56
Table 2: Serious Incidents and People Evacuated due to HAZMATS Spills.
Year Serious Incidents Total People Evacuated
1993 176 16,936
1994 347 18,382
1995 318 9,414
1996 311 19,130
1997 271 24,500
1998 132 8,211
1999 93 12,846
2000 116 20,630
2001 78 5,654
2002 98 38,505
2003 60 5,197
2004 77 10,910
2005 104 14,599
2006 125 20,238
2007 95 9,147
2008 77 9987
Figure 20: HAZMATS and Serious Incidents
HAZMATS Serious Incidents Reports
y = -15.424x + 285.98
R
2
= 0.5577
0
50
100
150
200
250
300
350
400
Year
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
Serious Incidents
Linear (Serious Incidents)
57
Figure 21: People Evacuated Due to HAZMATS Spills
Figure 22: Trends. Total spills, Serious Incidents, Total People Evacuated
y = 58.329x + 16207
R² = 0.014
y = -190x + 10731
R² = 0.0801
y = -152.7x + 15091
R² = 0.0028
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,000
45,000
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Total Spills
Serious Incidents
(*100)
Total People
Evacuated
58
Results Research Question One:
What is the Spatial Distribution of the Observed HAZMATS Spills?
H1: HAZMATS spills cluster together to create hotspots of toxic and hazardous incidents during
transport.
Maps and Visualization
Visualization of the results in a series of maps illustrate the spatial distributions of all
HAZMATS spills, severe spills, and evacuations were constructed both, at the national level and
for each of the nine major metropolitan regions. These visualization models were created with
the intent of revealing distribution patterns. In figures 2-14, it is evident that the majority of spills
tend to concentrate in densely populated urban areas.
Cluster Analysis, Unique Locations
A cluster analysis with a tolerance of 100 feet was used to identify “unique locations,”
and the number of spills per unique location was quantified. There were a total of 31,088 unique
identifiers or HAZMATS spills locations, with 30,700 identifiers reporting less than 100 spills in
total. The remaining 388 unique identifiers high frequency of spills (defined as 100 or higher)
HAZMAT discharges are listed and plotted in figure bellow.
59
Gettis-ord-gi * Analysis to Find Hotspots
A Gettis-ord-gi* hotspots analysis was run to find the HAZMAT facilities with high
frequencies of spills that were statistically significant. These facilities were located by using a Z
score larger than 1.29, or the 90
th
percentile, and were then identified as hotspots. Figure 23
below demonstrates the spatial distribution of hotspots at the national level. At this particular
scale, a threshold distance of 85 miles was utilized as suggested by the tool, since a distance of
stroke ½ miles cannot be processed for and extent as large as the Continental US with currently
available GIS programs and computing resources.
Figure 23: Map Unique Identifiers (HAZMATS spills) per Location, National Level
60
In Figure 23, it is noted that hotspots tend to occur in densely populated urban areas.
However, the scale of analysis does not facilitate detailed conclusions with regard to specific
subsets of the populations affected. An additional level of analysis is therefore needed to focus
on the areas where hotspots accumulate. Therefore, California is chosen as a case study,
discussed in the following section of this chapter.
Figure 24: Hotspots Analysis, 0.90 Percentile
California as a Case Study
Spatial Distribution of HAZMATS Spills in California
As previously stated, California has been chosen as a case study for a more careful
analysis. The 16,062 HAZMATS spills that occurred in California between the years 1998 and
2009 were selected to be used as case scenario. A cluster analysis with a tolerance of 100 feet
61
was used to identify “unique locations,” overlapping locations within the 100 feet bandwidth
were merged and consider as one location. The number of spills per unique location was
quantified as the “frequency of spills.” 354 unique locations were identified, with frequencies of
spills ranging from 1 to 1,288. Approximately two thirds of the 354 unique locations (239
locations) have a frequency of 20 or fewer discharges therefore are non-considered significant
polluters. The complete table, displaying the full data set containing this and other detailed
information is included in appendix V and VI.
This first visual model (Figure 25, below) shows that in general, HAZMATS spills and
Hotspots are clustered in two areas main areas, both of them highly populated and urban in
character: the first is in Northern California, and the second in Southern California.
62
Figure 25: California HAZMATS Spills, 1998-2009
Gettis-ord-gi* Hotspots Analysis for California as case study
A Gettis-ord-gi* hotspots analysis (with a 0.5 mile circular buffer around the hotspots)
was run to find the HAZMAT facilities with a statistically significant frequency of spills. The
threshold to define a location as statistical significance was kept consistent, defined as having a Z
63
score larger than 1.29, or the 90
th
percentile. Figure 26 below provides an illustration of the
spatial distribution of hotspots throughout the state of California.
Figure 26: Hotspots Analysis for California
The hotspots analysis shows a hotspots distribution pattern very similar to the pattern
shown by the frequency of spills. However, some abnormalities were detected; where some low
frequency polluters were identified as hotspots at the eastern section of the state. These points
64
may be considered to be false hotspots, as they were labeled as statistically significant polluters
in comparison with ‘all zero discharges neighbors’, in reality, none existing neighbors. This
phenomena is a limitation intrinsic to the method, a low frequency location can be falsely labeled
as a hotspot when it has a high frequency relative to its neighbors. In rural areas where there are
no other discharges within the half-mile radius, even a low-frequency (in absolute value) location
could become statistically significant when its neighbors are taken into consideration. The true
characterization of these points should be done through corroboration with local experts, either
by local surveys, interviews, focus groups with the local communities, or literature reviews that
use accessible local reports.
This validated hotspots analysis yielded 135 hotspots locations with a Z score of 1.29 or
greater. These locations have been incorporated into a GIS layer file using the GIS Spatial Join
function. Subsequently, the data were joined with the U.S. Census Bureau’s TIGER layer
entitled “County and County Subdivisions” for California. Spatial information attributes were
conferred, so the hotspots are geographically referenced with County Subdivisions, States and X-
Y. The complete list of locations is presented in Appendix IV.
In this study, the Nearest Neighbor Distance Analysis (NNDA) was used to obtain the
statistical parameters that describe the spatial distribution of the hotspots at the state level. This
Nearest Neighbor Ratio (NNR) yielded a 0.283917 value and a Z-score of -59.98; clearly
indicating that clustering of hotspots does exist in California, since NRR is less than one.
Statistical significance is granted by such an extreme Z score, which indicates that there
is less than 1% likelihood that this clustered pattern could be the result of random chance.
Having hotspots so highly clustered in certain areas is an extremely important finding. Several
HAZMAT facilities with high frequencies of spills lie within proximity of one another,
65
compounding the risk and becoming an exponential threat to the surrounding communities.
Additionally, these clusters are usually located within urban centers of high population density,
meaning that a significant number of people would be affected by a catastrophic event such as a
severe spill. Even in the best case scenario, this situation could lead to compounded health
hazards for the populations that are subjected to multiple HAZMAT spills sources. It is
extremely relevant that priority be given to these particular areas, so that policy changes and
HAZMATS best practices can be implemented.
The next section of this thesis details the characteristics of the populations living near
these hotspot clusters.
Results Research Question Two:
How Are HAZMATS Spills Distributed Across Social Groups?
H2: HAZMATS spills occur disproportionately in poor neighborhoods and in neighborhoods of
color.
To answer questions about what specific populations are affected by particular hotspots,
it is necessary to focus on a small geographic scale a small sample of hotspots clustered in
Southern California has been selected to be correlated with its affected population. Since
population characteristics vary widely, or are in fact, unique to each community, information on
communities should be provided using the highest level of detail available.
The population data were obtained at tract level resolution from the U.S. Census
Bureau’s America Fact Finder Summary File SF3, 2000. This was the most recent dataset
available, since data for the 2010 census was not yet publicized at the time this research was
conducted. For this section of the study, the variable of interest is the “percentage of people by
66
race,” since race and ethnicity are considered to be reliable indicators of belonging to a minority
(protected) population. Ten fields of information were selected, including the following: one
general demographic indicator (total population), seven indicators of race (percentage of Whites,
percentage of Blacks, percentage of Asians, percentage of Native Americans, percentage of
Pacific Islanders, percentage of other races, and percentage of two or more races), and two
indicators of ethnicity (percent of Latino of any race, percent of White Non-Latino or Hispanic).
The SF3 population tables were downloaded for each county, and a database
incorporating the 63 California counties was compiled to represent the population coverage at
the state level. This database was then overlaid with a geographic layer, which in turn made a
spatial join. Afterwards, the layer was joined with the U.S. Census Bureau’s TIGER layer
entitled “Tracts” for California, so that population data could be geographically referenced at the
tract level. This layer was labeled as Percentage of Population by Race at the tract level (PPR).
The hotspots layer is an event layer that contains a Z score and a frequency value
associated to a location (X, Y), that represent an area of a 0.5 mile radius. Then, another “buffer”
layer was created, buffering each location with a 0.5 mile fixed radius, allowing for the
dissolution of features where buffers intersect. This layer was called the Major Hotspots Areas
(MHA) layer (Figure 27).
67
Figure 27: Hotspots Areas in California
By clipping the Percentage of Population by Race (PPR) layer with the Major Hotspots
Areas (MHA) layer, it is possible to calculate the percentage of the population directly contained
within the major hotspots areas. For cases where the tract is not intersected in its totality by the
hotspot area, the tracts are clipped and their population attributes are arithmetically calculated to
include only the proportional part of the population’s information in order to prevent the
68
introduction of artificial boundaries by having to manage the totality of each covered the tract
area as a whole.
The consolidated racial composition of the MHA area is presented in Table 4.
Information is reported in terms of tract-average percentage of the population by race, by tract-
average actual population by race, and by several other basic statistical descriptors. The statistics
of the total population of California can be compared with the statistics of the population
contained within the Major Hotspots Areas. We notice that minorities account for more than
70% of the total population living in MHA areas, since the group identified as “White Non-
Latino reaches a peak of 29% of the total population in these areas. This is to be contrasted with
the California state average, where minorities account for about half the population (and White
Non-Latino represents 47%). See Table 5 for more information.
Table 3: Summary of Population within Major Hotspots Areas by Race and Ethnicity
Table 4: Summary of Total Population by Race in California
Statistics Population at HAS in CA
Total White Black Native Asian Pacific Other 2 or More
Any -
Latino
White non
Latino
Mean 1,633 790 105 18 188 5 452 75 829 473
Percent 100.0% 48.4% 6.4% 1.1% 11.5% 0.3% 27.7% 4.6% 50.8% 29.0%
Standard Error 103 58 19 1 24 0 39 5 66 46
Median 1,089 436 21 11 43 1 177 46 396 127
Stand Deviation 1,737 981 311 23 403 8 651 83 1,110 779
Maximum 9,984 6,493 3,754 136 3,847 50 3,844 478 7,074 6,175
Population Sum 460,550 222,707 29,653 5,166 53,088 1,359 127,527 21,053 233,843 133,512
Tracts Count 282 282 282 282 282 282 282 282 282 282
Statistics Total Population CA
Total White Black Native Asian Pacific Other 2 or More
Any -
Latino
White non
Latino
Mean 4,767 2,884 286 45 518 16 794 223 1,545 2,261
Percent 100.0% 60.5% 6.0% 0.9% 10.9% 0.3% 16.7% 4.7% 32.4% 47.4%
Standard Error 24 19 6 1 9 0 11 2 19 19
Median 4,550 2,707 102 35 257 8 414 203 955 2,117
Stand Deviation 2,018 1,560 540 45 719 30 912 129 1,561 1,612
Maximum 36,146 23,025 5,765 1,405 7,419 732 6,917 1,916 11,235 20,567
Population Sum 33,599,753 20,327,679 2,015,908 317,671 3,653,932 114,016 5,596,712 1,574,114 10,893,584 15,940,197
Tracts Count 7,049 7,049 7,049 7,049 7,049 7,049 7,049 7,049 7,049 7,049
69
A more detailed look at this comparison is included in Table 5, summarizing the
arithmetical difference between the percentage of each race population group in California
versus the percentage of each specific race population group within the HMA is calculated. We
will call this difference percentage of change. For any given group, a negative change indicates a
“less-than-usual” representation of this population within the HMA while a positive total change
indicates the presence of an atypically high percentage of this population group within the HMA.
(Less-than-usual or in Higher-than-usual in comparison with the representation of this population
group within the whole California)
The results of this study did not find all minorities over-represented in the Major
Hotspots Areas. Blacks, Native Americans, Asians, Pacific Islanders, and those who identify as
two or more races seem to have similar representation in both the MHA and the California state
level.
It is also important to account for how important is the change within a population by
measuring proportionality of the change within the population, this is, the percentage of change
relative to its own base value. This approach overcomes the fact that a relatively small change,
(such as 2%) represents a completely different process when it occurs within a petite minority
than when occurs in a large minority (such as 40%). For example in a population totaling 4% of
the sample, a 2% change is actually experiencing a swelling proportional increase of 50%, while
when the same conditions are applied to a group totaling the 40% of the sample, the certain total
change of 2% represents a proportional increase of a mere 5%. Table 5 demonstrates this point
by highlighting that proportional changes within the population have the same direction but are
sharper than their total change counterparts.
70
Table 5: Summary of Changes in Percentage of Population by Race
A very simple statistical tool called Test of Means, or the T-Test, was then used to
determine if the MHA and their effects are evenly or disproportionately distributed between
populations. The T-Test helps determine whether the mean value of a sample (its Observed
Value) is significantly different than the mean of the population (its Expected Value).
In this study, the mean values of population by race in California (Expected Values) are
compared with the mean values of population by race in the MHA (Observed Values). It is
expected that if the MHA is distributed uniformly over any population, the mean values for the
total population should be the same as the mean values for the population within the MHA.
The null hypothesis states that the difference between the Mean (Expected Value) minus
the Mean (Observed Value) is zero. The alpha threshold has been set to 0.05 percent to reject the
null hypothesis within 90 percent certainty. The test result is a number called “t-stat,” which is a
normalized statistic of the T-probabilistic distribution. “T-stat” indicates that the T statistic
threshold for the probability of falsely rejecting the null hypothesis is equal or lower than alpha.
Table 6 summarizes the T-test results for this study. The interpretation is straightforward. There
exists a box for every pair of variables among California and MHA datasets to which a T-test
was applied.
For instance, the first box is labeled “% White,” meaning that the box contains the result
of a T-test between the percentage of the total population that is identified as White in both
datasets. The first three rows in each box describe the input series with two columns: expected
(California) values on the left, and actual (MHA) values on the right. These columns display the
White Black Native Asian Pacific Other 2 or More
Any -
Latino
White non
Latino
Total change -12.1% 0.4% 0.2% 0.7% 0.0% 11.0% -0.1% 18.4% -18.5%
Proportional change -20.1% 7.3% 18.6% 6.0% -13.1% 66.2% -2.4% 56.6% -38.9%
71
mean, variance, and number of observations in the input series. Test results are shown in the
fourth row, displayed in single column format. After that, the table shows “hypothesized mean
difference” and “degrees of freedom” (df); the T-statistic threshold value and the corresponding
probability of a false null hypothesis rejection in one tail and two tail scenarios. This table also
shows the critical value for the T-statistic at an alpha probability.
Table 6: Summary Test of Means California Population by Race
The T-test results prove that the lower-than-usual percentage of Whites living in the
MHA has statistical significance, since P=0 (with two digits of precision). Conversely, the test
validates the finding that there are unusually high percentages of Latinos, Native Americans, and
other minorities living within the MHA.
t-Test: Two-Sample Assuming Unequal Variances
Total Population % White % Black % Native % Asian
CA MHA CA MHA CA MHA CA MHA CA MHA
Mean 4766.34 4745.43 60.57 48.47 6.04 6.40 0.94 1.10 10.53 11.05
Variance 4072600.12 4013966.90 494.60 426.39 122.30 105.56 0.79 0.39 166.19 262.76
Observation 7049 282 7049 282 7049 282 7049 282 7049 282
Test Results
Hypothesiz
df
t Stat
P(T<=t) on
t Critical on
P(T<=t) two
t Critical tw
% Pacific % Other Race % Two or More Race % Latino any Race % White non Latino
CA MHA CA MHA CA MHA CA MHA CA MHA
Mean 0.33 0.32 16.86 27.78 4.73 4.63 32.40 50.80 47.41 29.26
Variance 0.34 0.26 242.36 351.14 3.89 1.90 663.46 922.32 788.54 686.37
Observation 7049 282 7049 282 7049 282 7049 282 7049 282
Test Results
Hypothesiz
df
t Stat
P(T<=t) on
t Critical on
P(T<=t) two
t Critical tw
0.00
303.00 306.00 306.00 328.00 294.00
0.00 0.00 0.00 0.00
-1.57
0.43 0.00 0.34 0.00 0.06
0.17 10.39 0.43 -4.14
1.65
0.86 0.00 0.67 0.00 0.12
1.65 1.65 1.65 1.65
1.97
0.00 0.00 0.00 0.00 0.00
1.97 1.97 1.97 1.97
306.00
0.22 -10.08 -0.02 -10.44 12.36
310.00 296.00 328.00 296.00
0.00
1.65 1.65 1.65 1.65 1.65
0.41 0.00 0.49 0.00
0.00
1.97 1.97 1.97 1.97 1.97
0.82 0.00 0.98 0.00
72
Therefore, based on the results of this study, it can be concluded that for the Latino and
Native American communities, the incidence of HAZMAT hotspots is significantly higher than
what should be expected if the MHA affected all populations equally. These conclusions support
the hypothesis that HAZMAT spills contribute to environmental injustice by disproportionately
burdening neighborhoods of color.
Socioeconomic Characteristics of the Population within MHA
A similar analysis can be performed to assess the socioeconomic characteristics of the
tracts located within the MHA. For the purposes of this study, only an exploratory comparison of
the percentage of people living in poverty near the MHA has been performed. A visual analysis
of the intersection between the MHA and a layer depicting poverty levels suggests that the
poverty levels form concentric circles away from the centers of hotspots.
These findings indicate that there are a disproportionately high number of frequent or
major spills in socially vulnerable (poor) areas (Figure 28). It is strongly suggested that these are
places where environmental injustice is occurring, and are therefore worthy of further study. The
strong correlation displayed through this analysis, as well as the reference to this phenomenon in
general made by other authors, stresses the need for further studies.
73
Figure 28: Spatial Distribution of MHA and Poverty Levels
74
CHAPTER 5: CONCLUSIONS
This thesis has explored the relationship between high-frequency HAZMATS spill locations and
communities in close proximity to those locations that are characterized by social vulnerability
and /or high levels of poverty. This analysis was performed through the application of spatial
analysis techniques at two different scales, first at the Continental US, and then the selection of
the state of California as a case study. This study has identified a significant HAZMATS spill
data gap between the years 1991 and 1997. Work by other authors (Cutter, et. al 1997) has
identified decennial trends in HAZMATS distribution from 1971 to 1991. This study resumes
similar analysis from 1998 to 2009. If future research focuses on filling this gap in the research, a
forty year time-series-analysis could be accomplished.
This study concludes that the number of spills increased slightly yet steadily from 1998-
2006, most likely reflecting a normal increase in the amount of HAZMAT production. However,
in 2006 the introduction of new legislation significantly increased the monitoring of spill event
types and the level of enforcement. As a result, a significant peak of HAZMATS spills can be
observed after this legislation was passed. After that peak, the number of spills has declined. In
spite of the aforementioned increase in HAZMAT spill frequency; there is a steady decrease in
the number of serious incidents and in the number of people evacuated. This is perhaps
indicative of better practices adopted by industries that use HAZMAT materials.
The methodology applied in this study is tailored to meet the legal requirements of the
Environmental Protection Agency (EPA) in Environmental Justice Determinations (EJD). This
methodology also offers solutions for problems involving MUAP and the introduction of
artificial boundaries by using census tract boundaries by allowing the user to test several
75
parameters in a consistent way and determine the parameters and units of measure that are best
suited to a particular analysis.
Responding to the research question one, -What is the spatial distribution of the observed
HAZMATS spills? - This analysis has concluded that HAZMAT spills cluster together, and that
clustering occurs at national, regional, and local scales in densely populated urban areas
connected by mayor transportation highways confirming the research hypothesis H1:
HAZMATS spills cluster together to create “Hotspots” of toxic and hazardous incidents during
transport.
We have also shown that most hotspots of HAZMAT spill incidents take place within
these clusters. This clustering of HAZMAT hotspots facilities increases the risks posed to
surrounding communities.
In response to the research question two: -How are HAZMATS spills distributed among
social groups? - with H2: HAZMATS spills occur at a disproportionately higher level in poor
neighborhoods and in neighborhoods of color; by comparing populations within the identified
HAZMATS hotspots areas to the general population of California, it has been concluded that the
differences among the two samples are statistically significant. More specifically, major
differences were observed among the following groups: Whites and White Non-Latinos have a
lower-than-expected presence at Major Hotspots, but Latinos, Native Americans, and persons
who identify as another race have a presence within the HAZMATS hotspots areas that is
significantly greater than what one would expect if the MHA affected all populations equally.
Visualization of the results of this analysis suggests that for some areas in California, the
poverty levels form concentric circles away from the center of hotspots areas and creates a
concentration of poverty around the identified hotspots. This emerging pattern is comparable to a
76
“bull’s eye” pattern, and has been identified by previous authors such as Anderton (Anderton et
all 1994: 239). A pattern of this type implies the existence of inequity and warrants further
research. Findings of a disproportionately high occurrence of frequent or major spills on socially
vulnerable (poor) areas indicate that these are places where environmental injustice is occurring.
This topic also warrants further study in the future.
One more step is necessary to achieve the objectives of this thesis. Scientific findings
such as these, need to be transformed into policy and recommendations and incorporated into
everyday practice. Drawing connections about unfair exposure of socially vulnerable populations
to HAZMATS spills under the current legal framework of Environmental Justice, and Civil
Rights, while identifying explicit actions that public agencies charged with protecting these
populations can take according to their missions and responsibilities, is the only way sound
science can be transformed into enforceable policy.
In conclusion, the geographic regions identified as Major Hotspots Areas (MHAs) in this
study need to receive priority by policymakers, so that the EJ issues such as the ones discussed in
this thesis can be documented, addressed, and eventually resolved.
77
BIBLIOGRAPHY
• Anderton, D. L. in him et al., 1994. Environmental Equity: The Demographics of
Dumping. Demography 31(2):229-248
• Anderton, D.L., Anderson, A.B., Rossi, P.H., Oakes, J.M., Fraser, M.R., Weber, E.W., &
Calabrese, E.J. (1994). Hazardous waste facilities: Environmental equity issues in
metropolitan areas. Evaluation Review, 18, 123-140.
• Been, V. 1990. Locally Undesirable Land Uses in Minority Neighborhoods:
Disproportionate Sitting of Market Dynamics. The Yale Law Journal 10:1383.
• Boer, J. T. et al. 1997. Is there Environmental Racism? The Demographics of Hazardous
Waste in Los Angeles County. Social Science Quarterly. 78(4):793-829.
• Boone, C G. November 01, 1999. Creating a Toxic Neighborhood in Los Angeles
County: A Historical Examination of Environmental Inequity. Urban affairs review
(Thousand Oaks, Calif.) 35, no. 2, (accessed March 24, 2011)
• Brooks, N. and R. Sethi. 1997. The Distribution of Pollution: Community Characteristics
and Exposure to Air Toxics. Journal of Environmental Economics and Management
32:233-250.
• Bullard R.D. 1996. Environmental Justice: It’s More Than Waste Facility Sitting. Social
Science Quarterly 77(3): 493-449.
• Bullard, R. D. 1994. Dumping in Dixie: Race, Class, and Environmental Quality: Second
Edition. Boulder: Westview Press.
• Bullard, R.D. (1990). Dumping in Dixie: Race, class and environmental quality. Boulder,
CO: Westview Press.
• Burke, LM. (1993). Race and environmental equity: A geographic analysis in Los
Angeles. Geo-Info Systems October, 44-50.
• CEQ, “Environmental Justice Guidance Under the National Environmental Policy Act,”
(December 10, 1997), Attachment 1 (copy of Interagency Working Group guidance
under E.O. 12898).
• Council on Environmental Quality (CEQ). (1971). The second annual report of the
council on environmental quality. Washington, DC: U.S. Government Printing Office.
78
• Cutter, S. L. and J. Minhe. 1997. Trends in U.S. Hazardous Materials Transportation
Spills. The Professional Geographer 49, no. 3 (1997): 318; 0033-0124
• Cutter, S. L., D. Holm, and L. Clark. 1996. The Role of Geography Scale in Monitoring
Environmental Justice. Risk Analysis 16(4):517-526
• EPA 1995c. Federal Register 60: 7824 (from Feng 2001)
• General Accounting Office (GAO). (1983). Siting of hazardous waste landfills and their
correlation with racial and economic status of surrounding communities. Washington
DC: General Accounting Office.
• Glickman, T. S., D. Golding, and R. Hersh. 1995. GIS-based Environmental Equity
Analysis: A Case Study of TRI Facilities in the Pittsburg Area, pp95-114 in Computer
Supported Risk Management, edited by W. A. Wallace and E. G. Beroggi. Dordrecht, the
Netherlands: Kluwer.
• Greer, Edward. 1995. Andrew Hurley, Environmental Inequalities: Class, Race and
Industrial Pollution in Gary, Indiana, 1945-1980 (Book Review). Socialism and
democracy 19, (accessed March 24, 2011).
• Hersh, R. 1995. Race and the Industrial Hazards: A Historical Geography of the
Pittsburgh Region, 1900 to 1990. Resources for Future Discussion Paper, 1995 (95-18.)
Technical Report.
http://www.epa.gov/compliance/ej/resources/policy/ej_guidance_nepa_epa0498.pdf
• Iseki, H. June 01, 2001. Environmental justice: Spatial distribution of hazardous waste
treatment, storage and disposal facilities in Los Angeles. Journal of urban planning and
development 127, no. 2, (accessed March 24, 2011).
• Lee, C. (ed.) 1992. Proceedings of the First National People of Color Environmental
Leadership Summit. New York: United Church of Christ Commission for Racial Justice.
• Mohai, P. and B. Bryant. 1992. Environmental Racism: Reviewing the Evidence, pp.
163-176 in Race and Incidence of Environmental Hazards: A time for Discourse, edited
by B. Bryant, and P. Mohai, Boulder, Colorado: Westview Press
• NCHRP Project 8-36: Technical Methods to Support Analysis of Environmental Justice
Issues. Final report of project NCHRP Project 8-36(11). Cambridge Systematics, Inc.
79
(2002) Transportation Research Board, National Research Council. Washington, DC:
National Academy Press.
• NCHRP Report 532: Effective Methods for Environmental Justice Assessment,
Forkenbrock and Sheeley (2004)
• Pastor, Manuel. January 01, 2001. Which Came First? Toxic Facilities, Minority Move
In, and Environmental Justice. Journal of urban affairs 23, no. 1, (accessed March 24,
2011).
• Perlin, S. A. et al. 1995. Distribution of Industrial Air Emissions by Income and Race in
the United States: An Approach Using the Toxic Release Inventory. Environmental
Science and Technology 29 (1):69-80.
• Phillips, Carl and Sexton, K. (1999). Science and policy implications of defining
environmental justice. Journal of exposure analysis and environmental epidemiology 9,
no. 1, (accessed March 24, 2011).
• Pollock, III, P. H. and M. E. Vittas 1995. Who Bears the Burden of Environmental
Pollution? Race, Ethnicity and Equity in Florida. Social Science Quarterly 76(2):294-
310.
• Ringquist, E. J. 1997. Equity and the Distribution of Environmental Risk: The Case of
TRI Facilities. Social Science Quarterly 78(4): 811-829.
• Schnaiberg, Allan, 2001. The Environmental Justice Movement: Equitable Allocation of
the Costs and Benefits of Environmental Management Outcomes. Social justice research
14, no. 4
• Schweitzer, Lisa (2004). "Environmental sacrifice zones: Risk and transport in southern
California". Ph.D. diss., University of California, Los Angeles, 2004. In Thesiss &
Theses: Full Text [database on-line]; available from.
• Schweitzer, Lisa. 2008. No Notice Evacuations, Urban Form and Environmental
Injustice: An Exploratory Study. Research Proposal Submitted to METRANS
• Schweitzer, Lisa. November 01, 2006. Environmental justice and hazmat transport: A
spatial analysis in southern California. Transportation research. Part D, Transport and
environment 11, no. 6, (accessed March 24, 2011).
80
• Shaw, G. et al., 1986. Characteristics of hazardous material spills from reporting systems
in California. Am J Public Health. 1986 May; 76(5): 540–543. PMCID: PMC1646624
• Sui, D. Z. and J. R. Giardiano. 1995. Applications of GIS in Environmental Equity
Analysis: A Multi-Scale and Multi-Zoning Scheme Study for the City of Houston, Texas,
USA. pp. 950-959 in Proceedings GIS/ LIS’95.
• Talih, M., & Fricker, R.D. (2002). Effects of neighborhood demographic shifts on
findings of environmental injustice: a New York City case-study. Journal of the Royal
Statististical Society. Series A, 165(2), 375-397.
• Tobler, W. 1979. Cellular Geography, pp. 379-386 in Philosophy in Geography, S. Gale
and G. Olson, (eds.) Dordrecht: Reidel. (in Feng 2001)
• U.S. EPA. 1998 Final Guidance for Incorporating Environmental Justice Concerns in
EPA’s NEPA Compliance Analysis. EPA Office of Federal Activities, Washington, DC.
• United Church of Christ Commission for Racial Justice (UCC). (1987). Toxic wastes and
race in the United States: A national report on the racial and socio-economic
characteristics of communities with hazardous waste sites. New York, NY: Public Data
Access Inc.
• United States Environmental Protection Agency (EPA). (1998). Final Guidance For
Incorporating Environmental Justice Concerns in EPA’s NEPA Compliance Analyses.
Retrieved March 11, 2009.
• Wang, Y. January 01, 2004. Environmental justice: frequency and severity of US
chemical industry accidents and the socioeconomic status of surrounding communities.
Journal of epidemiology and community health (1979) 58, no. 1, (accessed March 24,
2011).
• Werner, R.J. 1997. Toxic Releases and Demography in Minneapolis/ St Paul: A GIS
Exploration. Proceedings from the 1998 Environmental Systems Research Institute Users
Conference.
• Whitney, Mike. "The Siege of New Orleans.” The Gaither Reporter (2005): 26. Ethnic
News Watch (ENW), Pro Quest. Web. 17 Sep. 2011.
81
• Wrigley, N. et al. 1996. Analyzing, Modeling and Resolving the Ecological Fallacy, pp.
23-40 in Spatial Analysis: Modeling in a GIS Environment, P. Longley and M. Batty,
(eds.) Cambridge, England: Geo-Information International.
82
APPENDICES
Appendix I: Inventory of Federal Portals Containing Databases and
other Information relevant to HAZMATS Spills
• Toxics Release Inventory Program (TRI) - The Toxics Release Inventory (TRI) is a
publicly available database containing information on toxic chemical releases and waste
management activities reported annually by certain industries as well as federal facilities.
• The Comprehensive Environmental Response, Compensation, and Liability
Information System (CERCLIS) contain information on hazardous waste sites, site
inspections, preliminary assessments, and remediation of hazardous waste sites.
• Hazardous Waste Data - Access to information from the Resource Conservation and
Recovery Act Information System (RCRA Info).
• National Response Center - Serves as the sole national point of contact for reporting all
oil, chemical, radiological, biological, and etiological discharges into the environment
anywhere in the United States and its territories,
• RCRA Online Identifies and indexes over 2900 RCRA letters, memoranda, and
Questions and Answers.
• Integrated Risk Information System (IRIS) - A database of human health effects that
may result from exposure to various substances found in the environment.
• Emergency Response and Notification System (ERNS) would be the Primary source of
HAZMATS spills reports. The Emergency Response Notification System (ERNS) is a
database containing notifications of oil discharges and hazardous substances releases.
The ERNS program is a cooperative data sharing effort among the Environmental
Protection Agency (EPA), the Department of Transportation DOT and the National
Response Center (NRC). The types of release reports contained into this database are
substances designated as hazardous substances under the Comprehensive Environmental
Response, Compensation, and Liability Act of 1980 (CERCLA), petroleum products, as
defined by the Clean Water Act of 1972 (CWA), and other types Hazardous materials.
• HAZMATS Information System (HMIRS) Central repository U.S. Government
military services and civil agencies storing Material Safety Data Sheets (MSDS) , Hazard
Communications Standards (HAZCOM) warning labels and Department of
Transportation information. HMIRS provides this data for hazardous materials purchased
by the Federal Government and assists its personnel who handle, store, transport, use, or
dispose of hazardous materials.
83
• Hazardous Substances and Emergency Events Surveillance (HSEES) this database
was established by the Agency for Toxic Substances and Disease Registry (ATSDR) to
collect and analyze information about acute releases of hazardous substances and
threatened releases that result in a public health action such as an evacuation. Only
fourteen state health departments currently have cooperative agreements with ATSDR to
participate in HSEES: Colorado, Florida, Iowa, Louisiana, Michigan, Minnesota, New
Jersey, New York, North Carolina, Oregon, Texas, Utah, Washington, and Wisconsin.
84
Appendix II: Legal Framework of Environmental Justice
In addition to understanding the evolution of EJ as a discipline, it is important to
summarize and understand the various statutes and regulations that establish the legal and
philosophical underpinnings of current EJ practices. This allows any new advances in knowledge
to be easily incorporated into EJ practice.
Chapter 21, Civil Rights Act (1964) allows a private plaintiff to bring claims of disparate
environmental impact burdens, which gives standing to the average citizen to enforce EJ
principles and to oversee their implementation.
Title VI, Civil Rights Act (1964) states that any program or activity receiving federal
financial assistance shall remain free of discrimination by race, color, gender, or national origin.
Any project or program that exhibits or results in discriminatory practices could lose federal
funding.
The National Environmental Policy Act (NEPA) (1969) requires federal agencies to
consider environmental impacts before taking actions that could significantly affect the
environment. NEPA establishes the protocols and procedures needed to evaluate all
environmental impacts of a proposed project, and also evaluates the effects of these impacts
across different sectors of the population.
The intent of NEPA is to provide access and opportunities for public participation to all
sectors of the population, throughout the entire environmental review process, without any bias
or discrimination. This ensures that any project resulting in discriminatory environmental
practices shall conclude with a negative recommendation.
Under NEPA, all “reasonably foreseeable effects” associated with the project under
review need to be considered and clearly identified in an Environmental Impact Study (EIS)
report. The EIS report should emphasize the impacts of the project on all social groups affected,
including those groups protected under a special status such as minority and ethnic populations,
the elderly, handicapped, and/or any other socially vulnerable population as denoted by specific
Executive Orders.
The process of Environmental Impact Review needs to address specific questions of
proportionality regarding the distribution of environmental benefits and burdens among different
sectors of the population. The process shall also identify possible mitigation measures in order to
85
minimize environmental impacts, especially if these impacts impose a disproportionate burden
on protected populations.
However, it is important to note that NEPA cannot guarantee either the enforcement or
establishment of specific rights, legal recourse, or standards for the protection of any specific
population group. Since NEPA is a “procedural statute,” it cannot impose specific outcomes on a
given project. NEPA cannot stop a project, even if the findings of an EIS report demonstrate
disproportional impacts over a particular resource or population group. These instances are
classified as “findings of no practicable alternative.”
Statutes Protecting Vulnerable Populations:
Title VIII of the Civil Rights Act (1968) is also known as the “Fair Housing Act." This
legislation is triggered when a proposed project would either displace vulnerable populations or
affect their place of residence. The Fair Housing Act is intended to prevent discrimination
practices in the housing market.
The Age Discrimination Act of 1975 prohibits age discrimination in all federal agencies
and in all federally-assisted programs and initiatives.
The Americans with Disabilities Act (1990) protects persons with disabilities, and the
Rehabilitation Act of 1973 protects individuals with permanent handicaps. The Rehabilitation
Act of 1973 ensures that “no qualified handicapped person in the United States shall, solely by
reason of his handicap, be excluded from participation in, be denied the benefits of, or be
subjected to discrimination under any program or activity receiving federal financial assistance.”
Executive Orders (Implementation and Enforcement)
In addition to the aforementioned laws, the executive branch of the federal government
has enacted several Executive Orders (EOs) related to EJ issues. While Executive Orders are
policy and not law, they are nonetheless important since these orders are issued directly by the
Office of the United States President. The executive branch monitors several aspects of all
Federal Departments and Agencies’ operations, including budget approval. Failure to adhere to
EOs can result in serious consequences for noncompliant agencies, including the veto of their
operational budget.
86
The following three EOs are pertinent to the field of EJ: Executive Order 12898,
Executive Order 13166, and Executive Order 13045.
Executive Order 12898: “Federal Actions to Address Environmental Justice in Minority
Populations and Low-Income Populations.” EO 12898, issued in 1994 by President Bill Clinton,
requires federal agencies to make EJ part of their daily mission by evaluating and preventing any
disproportionate burdens on human health or the environment. EO 12898 sets the legal standard
for federal agencies in order to fulfill the requirements of the two cornerstone pieces of EJ
legislation: Title VI and NEPA.
Executive Order 13166 requires all federal agencies to make federally conducted
programs and activities more accessible to eligible persons with limited English proficiency.
Executive Order 13045 requires all federal agencies to identify and assess
environmental health risks and safety risks that may disproportionately affect children. As a
result of EO 13045, children became another protected population with regard to EJ issues.
Environmental Regulation and Authorities Specific to HAZMATS
This section lists the main agencies, laws, and amendments that regulate HAZMATS, for
the purpose of providing a solid background and highlighting some important considerations.
However, it should be emphasized that this list is by no means complete.
The US Environmental Protection Agency (EPA) is the principal agency responsible for
issues related to human health and the environment. Within the EPA, the Office of Solid Waste
and Emergency Response (OSWER) is responsible for regulating HAZMATS and overseeing
the proper functioning of HAZMATS handling facilities. OSWER applies four federal statutes to
help guide regulation, control of violations, general issues of implementation, liability,
remediation, and compensation for incidents related to HAZMATS management. These four
federal statutes are listed as follows:
1) Resources Conservation and Recovery Act (RCRA);
2) Toxic Substances Control Act (TSCA);
3) Federal Insecticide, Fungicide and Rodenticide Act (FIFRA); and
4) Federal Comprehensive Environmental Response, Compensation and Liability Act
(CERCLA, or “superfund”).
87
The Resource Conservation and Recovery Act (RCRA), [42 U.S.C. §6901 et seq.
(1976),] gives the EPA the authority to control hazardous waste from "cradle-to-grave.” The
EPA traces hazardous waste over the course of its life including its generation, transportation,
treatment, storage, trade, use, disposal, and collection to its final destination. The Federal
Hazardous and Solid Waste Amendments (HSWA), made to RCRA in 1984, are focused on
hazardous waste minimization and the phasing-out of HAZMATS landfills. These amendments
also allow for strong corrective actions over past or current releases of hazardous materials.
The Toxic Substances Control Act of 1976 (TSCA) grants the EPA authority to require
reporting, recordkeeping, testing requirements, and use restrictions for all hazardous chemical
substances and/or mixtures throughout the United States. TSCA Sections 12(b) and 13 also
require those who import or export hazardous chemicals to comply with certification reporting
and/or other requirements imposed by the EPA.
In Section 4, TSCA provides guidelines and regulations for chemical testing by
manufacturers, importers, and processors, when risks of HAZMATS exposures are either found
or suspected.
Section 5 of TSCA, the “Issue Significant New Use Rules (SNURs),” controls any new
HAZMATS or the "significant new use" of HAZMATS that could result in exposures to, or
releases of, a substance of concern.
Section 8 of TSCA contains a HAZMATS inventory of more than 83,000 chemicals. As
new chemicals are commercially manufactured or imported, they are placed on the list. Section 8
also lists the reporting and recordkeeping requirements for those people who manufacture,
import, process, and/or distribute chemical substances. Furthermore, Section 8(e) invites any
person dealing with hazardous materials under any circumstance to file a voluntary incident
report.
The EPA treats all TSCA Section 8(e) submissions as voluntary "For Your Information"
(FYI) submissions. These submissions are not required by law, but are instead submitted by
industries and public interest groups for a variety of reasons. These FYI submissions have also
been incorporated into our records, along with the mandatory records located in the federal
databases, and are cited as the main sources of information for this thesis.
88
The Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA) provides for
federal regulation regarding the distribution, sale, use, and disposal of pesticides. The principle
objective of FIFRA legislation is to ensure that pesticides "will not generally cause
unreasonable adverse effects on the environment, defining unreasonable adverse effects.” This
is accomplished by taking into account the economic, social, and environmental costs of using
any pesticides, as well as analyzing the way in which these costs are distributed throughout the
population. This legislation is crucial, as pesticides comprise a large percentage of the total
volume of commercial HAZMATS.
The Comprehensive Environmental Response, Compensation, and Liability Act
(CERCLA/Superfund) establishes a Federal "Superfund" to clean up uncontrolled or
abandoned hazardous waste sites as well as accidents, spills, and other emergency releases of
hazardous pollutants to the environment.
Under the Superfund, the EPA is authorized to conduct site identification, monitor sites,
and respond to activities located at in any state. State actions in this respect are coordinated
through each state’s environmental protection agency or their corresponding waste management
agencies. The EPA has the authority to seek out parties responsible for any release and assure
that these parties cooperate in the cleanup.
Due to budget constraints, Superfund cleanups were temporarily de-authorized. However,
EJ advocates successfully reversed this ban in the mid-1980s. The Superfund Amendments and
Reauthorization Act (SARA) of 1986 reauthorized CERCLA to continue cleanup activities. In
addition, some technical requirements were added to the legislation to support the public’s right
to know. Title III, also known as the Emergency Planning and Community Right-to-Know Act
(EPCRA), guarantees the public’s right to know and makes pertinent Hazardous Material
information available to the general public. This information was used in the development of this
thesis.
Federal Databases Containing HAZMATS Spills Information
An inventory of federal databases containing HAZMATS-related information is included
in Appendix I. For this thesis, only databases containing HAZMATS spills information were
reviewed, and four major data sets were compiled with records dating from 1998 to 2009.
The three major sources of information compiled for this thesis include the Emergency
Response and Notification System (ERNS), used as the primary source of HAZMATS spills; the
89
HAZMATS Information System (HMIRS); and the Hazardous Substances and Emergency
Events Surveillance (HSEES) project. These three databases were used in conjunction to
document HAZMATS spills, while data from the U.S. Census were used to document
characteristics of populations residing next to those HAZMATS facilities that had experienced
significant spills.
90
Appendix III: Types of Units Commonly Used in EJ Analysis
• Legal Units: State, County, City, Incorporated Places, Metropolitan Areas, etc.;
• Administrative units: ZIP Codes;
• Geographic Information Systems Units: Point, Polygon, Buffer, Cluster;
• Statistical Entities: Statistical Metropolitan Areas (SMAs);
• Census Units: Blocks, Block Groups, and Census Tracts.
Blocks are the smallest unit expressed through the census and are bounded on all sides by
discernible features, such as cities, towns, county limits, highways, or streets. Blocks never cross
greater hierarchical boundaries such as city, county, or state boundaries, and range from 3,000
square feet (0.69 acres) to 40,000 square feet (0.92 acre) in area. Block Groups are clusters of
blocks, most typically 90 blocks. Census tracts are compact contiguous parcels of land accessible
by road. The boundaries of census tracts never cross county lines, and census tracts encompass
an average of 4,000 persons and 1,600 housing units.
Each unit is accompanied by a series of advantages and disadvantages. Census tracts are
usually an appropriate scale to represent neighborhoods, as they have well-defined boundaries
and do not experience a great deal of change from one decennial census to the next. From a
scientific perspective, this provides an ideal opportunity for comparison, as only minor
adjustments need to be made in order to conduct a legitimate analysis. Census tracts are smaller
in area than ZIP codes and result in fewer aggregation errors. In general, the impacts of the
project or phenomenon under investigation are usually larger than the census tract, and therefore
the need for a smaller unit of analysis is difficult to justify.
However, there are also disadvantages associated with the use of census tracts as a unit of
analysis. The main issue associated with the use of census tracts involves the assumption of
homogeneity. When collecting data at the tract level, it is assumed that the population inside the
respective boundary is homogeneous; however, for some tracts this assumption is untrue, thereby
leading to incorrect conclusions and lessening the importance of minorities within certain census
tracts. Bullard (1994) and Liu (2001) found pockets of minorities and low-income communities
that were disproportionally affected by environmental hazards, but were embedded in census
tracts that consisted primarily of non-minorities.
91
Furthermore, census tracts present a disadvantage when used as a primary unit of
analysis, as tracts can vary widely in size and area. A census tract includes anywhere between
3,500 and 8,000 persons, as well as between 600 and 3,200 housing units. Such variations in size
occur as a result of the U.S. Census Bureau’s parameters for maximum and minimum size.
Census tracts do not always cover rural areas where environmental hazards can occur.
The advantages of selecting ZIP Codes as the unit of analysis are much different than the
advantages of using census tracts. ZIP codes have been successfully used in marketing to
represent, describe, and predict population trends, demographic information, and socioeconomic
characteristics. ZIP Codes are more inclusive than census tracts at the rural level. However, the
use of ZIP codes is also associated with several disadvantages. A typical ZIP Code is at least
eight times larger than a typical census tract, and as a result ZIP Code analysis introduces more
aggregation errors when counting populations. However, the main disadvantage associated with
ZIP Codes is a lack of consistency over the course of time, as ZIP Codes’ boundaries often
change in order to accommodate the needs of mail distribution. ZIP codes vary tremendously in
size and exhibit different characteristics.
Several authors defend the Linear Distance Buffer as the most advantageous area unit.
For example, a circular buffer with a half-mile radius is similar in area to an average census tract.
There is a certain degree of consensus among academics that 0.5 miles is a distance that
intuitively connotes “proximity” (Glickman, Goldin, & Hersh (1995) in Feng 2001). A distance
of 0.5 miles has proven to be adequate as a minimum impact distance for a proper impacts
assessment dispersion model (Werner 1997 in Feng 2001). However, this metric is also plagued
with disadvantages. Previous studies have demonstrated that using a circle with a fixed radius as
a unit of analysis can significantly affect the results of the analysis, and variations in the radii of
analyses can lead to different statistical findings (Sur and Giardiano 1995). The appropriate level
of analysis will depend ultimately on the problem at hand and the area of impact, as
demonstrated in Table following table 2.
92
Table 2: Correct Unit of Analysis (Source: Sur and Giardiano 1995 in Feng 2001)
For Impact Area Circular
Buffer of
The Probability of Census
Tract Being the Right
Choice as a Unit of Analysis
The Probability of ZIP Codes
Being the Right Choice of Unit
of Analysis
0.5 mi to 3.0 mi 47 (%) 36.5 (%)
0.5 mi to 2.0 mi 40 (%) 21 (%)
2.0 mi to 4.0 mi 11 (%) 32 (%)
The second problem to consider involves the introduction of artificial boundaries. Phenomena
including fires, earthquakes, and hazardous material spills spread independently of ZIP codes,
census tracts, or any other boundary that is artificially drawn. During analysis, census tracts
could arbitrarily divide a continuous space, not allowing gradual representation of change and
further introducing blurriness in the representation of the sample.
Authors such as Schweitzer (2004) noticed that most EJ models dealing with HAZMATS
had not measured facility concentration or release concentration independent of census
boundaries. Schweitzer pointed out that these artificial boundaries often distort the phenomenon
that is under investigation. Risk and the effects of HAZMATS exposure are difficult to model.
Often, there is not a precise linear relationship between HAZMATS exposure and facility
proximity; rather, the relationship varies depending on chemical composition and the quantities
of chemicals released. Schweitzer (2004) proposes four cardinal methodological improvements
in order to measure facility and release concentrations independently of census boundaries:
1) Improved risk and nuisance identification on space;
2) A spatial methodology for identifying facility and transportation risk clusters;
3) A conceptual model of HAZMATS risk for freight transport; and
4) A method for generalizing risk with limited data.
This thesis supplements and expands upon Schweitzer’s research. Specifically, it proposes a
spatial methodology that finds geographic correlations between vulnerable populations and
HAZMATS incidents independent of census tract boundaries; this is accomplished by allowing
the model to calculate population through arithmetic proportions when a tract should not be fully
included in the analysis. Tract-level information will be used to assess the proportionality of
population distributions adjacent to HAZMATS facilities.
93
What is an appropriate unit of analysis? An ideal unit is readily available, is comparable,
maintains relevance over time, and is adequate to the scale of the phenomenon being analyzed.
An ideal unit introduces the least level of error by its variation and does not contain so much
uncertainty as to confound the results of the analysis. This unit of analysis will not necessarily be
the same for all studies, but it will certainly be comparable.
94
Appendix IV:Definition of Key Terms and Operational Definitions
• Polluter: HAZMAT facility identified as a “Unique Location”
• Unique Location: All events reported within a distance of 100 feet from each other, and
can thereby be attributed to the same polluter
• Frequency: Number of spills reported by unique location
• Hotspot: Unique location with a statistically significant high frequency of spills
• Hotspot Vicinity (HSV): Area surrounding the hotspot used to evaluate its statistical
significance. The HSV shape can either be a circular buffer of a fixed radius surrounding
the Hotspot, or it can be defined in terms of its distance to the nearest neighbor.
• Statistical Significance: For the purpose of this study, a threshold value of Z = 1.29 has
been pre-established in order to declare a polluter a “Hotspot.” Z = 1.29 is equivalent to a
percentile of 0.90, confirming that our point is a hotspot with a 90 percent of certainty. Z
= 1.29 ensures that at least 9 out of 10 locations will have less discharges than a point
labeled as a Hotspot.
• High level of poverty: Census tract with 25% or more of its population living below the
poverty level. The precise definition of poverty level is dependent on the scope of the
analysis, as different agencies define poverty using different criteria.
• Minorities Tract: Census tract with 25% or more of its population belonging to minority
ethnic groups.
95
Appendix V: Table of Nation-wide Unique Locations with annual
Frequency of Spills >100
96
Table of Nation-wide Unique Locations with annual Frequency of spills larger than 100
continued
97
Appendix VI: Table Hotspots with Z score =>1.29 or 0.90 Percentile
for California
98
Table Hotspots Z Score 1.29 or .90 Percentile for California Continued
99
Table Hotspots Z Score 1.29 or .90 Percentile for California Continued
Abstract (if available)
Abstract
This study encompasses Hazardous Materials Spills (HAZMATS) occurred in the United States between 1998 and 2009, also concentrating on California as a case study. This work expands the base of empirical knowledge and observations of previous studies dealing with the geographical location of Hazardous Materials Handling and Storage Facilities (HMHSF). ❧ Integrated into the HAZMATS spills distribution analysis, a frontline approach is proposed to address the more common methodological issues and constraints faced by the environmental justice discipline, as reported by previous authors. This new methodology is purposely designed to comply with the current legal requirements of the Environmental Protection Agency (EPA) for Environmental Justice Determinations (EJD), so can be expanded from theoretical to the applied realm. Two major methodological problems were identified during the literature review: a) The precise determination of the best geographical extent (scale) to define the area of study and the appropriate unit of analysis to capture the environmental problem at hand, and b) the introduction of artificial boundaries caused by the use of population census data aggregated at the tract, block, or ZIP code level. The author proposes, creates, and tests an innovative Geospatial Tool, based in the study of "Hotspots"—or outliers—and the use of a variable radius in order to address the "Modifiable Area Unit" problem and the introduction of artificial boundaries by the census data offering a solution for the two more common methodological issues that have been encountered by previous authors. ❧ This thesis also explores distribution patterns and spatial relationships between the more significant HAZMATS spills locations ("Hotspots") and highlights communities with high levels of poverty and social vulnerability where questions concerning environmental justice need to be urgently addressed. Future work is proposed to assess the presence of "Hotspots" as predictors of future severe spills and catastrophic events.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Resilience by anticipating change: simple and robust decision making for coastal adaptation planners, communities and elected officials
PDF
Roadway hazard analysis: a safe ride for motorcycles
PDF
Analysis of future land use conflict with volcanic hazard zones: Mount Rainier, Washington
PDF
Environmental justice in real estate, public services, and policy
PDF
The role of social media for community-based organizations focused on environmental justice in southeast Los Angeles
PDF
The interplay of environmental justice and environmental psychology in environmental conflicts: a case study of the La Farfana water sewage treatment plant and the Pudahuel community in Santiago Chile
PDF
Defining the geospatial needs of a ubiquitous disaster relief organization
PDF
A spatiotemporal analysis of racial disparity in the distribution of superfund sites within Santa Clara County, California
PDF
Risk analysis and assessment of non‐ductile concrete buildings in Los Angeles County using HAZUS‐MH
PDF
Location of warehouses and environmental justice: Three essays
PDF
Environmental equity and urban sustainability: an analysis of untreated household wastewater in Tijuana, Mexico
PDF
Integration and validation of deterministic earthquake simulations in probabilistic seismic hazard analysis
PDF
Semi-automated visualization of spatial information in unstructured text
PDF
California's adaptation to sea level rise: incorporating environmental justice communities along the California coastline
PDF
GeosocialFootprint (2103): social media location privacy Web map
PDF
The modifiable areal unit problem (MAUP) via cluster analysis methodologies: a look at scale, zoning, and instances of foreclosure in Los Angeles County
PDF
Finding food deserts: a study of food access measures in the Phoenix-Mesa urban area
PDF
Geospatial analysis of unintended casualties during combat training: Fort Drum, New York
PDF
A spatiotemporal analysis of environmental risk factors of Lyme disease in the Northeastern United States
PDF
Urban green space accessibility and environmental justice: a GIS-based analysis in the city of Phoenix, Arizona
Asset Metadata
Creator
Avendano, Claudia E.
(author)
Core Title
Environmental justice: geospatial impacts of hazardous materials spills
School
College of Letters, Arts and Sciences
Degree
Master of Science
Degree Program
Geography
Publication Date
05/14/2014
Defense Date
05/14/2014
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
environmental justice,geospatial,hazardous materials,HAZMATS,hotspots,OAI-PMH Harvest,vulnerable populations
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Schweitzer, Lisa (
committee chair
), Swift, Jennifer N. (
committee member
), Wilson, John P. (
committee member
)
Creator Email
cavendan@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-413632
Unique identifier
UC11297034
Identifier
etd-AvendanoCl-2513.pdf (filename),usctheses-c3-413632 (legacy record id)
Legacy Identifier
etd-AvendanoCl-2513.pdf
Dmrecord
413632
Document Type
Thesis
Format
application/pdf (imt)
Rights
Avendano, Claudia E.
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
environmental justice
geospatial
hazardous materials
HAZMATS
hotspots
vulnerable populations