NMPB*
Office of
Research and Development
Center for Public Health and
Environmental Assessment
Public Health and Environmental
Systems Division
vvEPA
United States
Environmental Protection
Agency
EPA/600/R-20/367 December 2020 www.epa.gov/ord
Environmental
Quality Index
2006-2010
Technical Report
1 WW
-------
-------
v>EPA
EPA/600/R-20/367
United States
Environmental Protection
Agency
ENVIRONMENTAL QUALITY INDEX
2006-2010,
Technical Report
Public Health and Environmental
Systems Division
Epidemiology Branch
Chapel Hill, NC
Office of Research and Development
Public Health and Environmental Systems Division
-------
Acknowledgments
Project Personnel
Danelle T. Lobdell, United States Environmental Protection Agency (EPA), Office of Research and Development (ORD), Center for
Public Health and Environmental Assessment (CPHEA)
Kristen M. Rappazzo, EPA, ORD, CPHEA
Stephanie DeFlorio-Barker, EPA, ORD, CPHEA
Alison K. Krajewski, Oak Ridge Institute for Science and Education (ORISE) Postdoctoral Grantee
Lynne C. Messer, Oregon Health and Science University-Portland State University School of Public Health, Support Contractor
Jyotsna S. Jagai, University of Illinois at Chicago, Support Contractor
Christine L. Gray, Duke University, ORISE Grantee
Monica P. Jimenez, Oak Ridge Associated Universities (ORAU) Student Services Contractor
Achal Patel, ORAU Student Services Contractor
Barbara Rosenbaum, General Dynamics Information Technology/Woolpert, Inc. (GDIT/Woolpert), Geographic information systems
(GIS) Contractor Support
Steven Jett, GDIT/Woolpert, GIS Contractor Support
External Peer Reviewers
Sheryl Magzamen, Department of Environmental and Radiological Health Sciences at Colorado State University's College of
Veterinary Medicine and Biomedical Sciences.
Anne M. Roubal, County Health Rankings & Roadmap at University of Wisconsin Population Health Institute
Ying Zhou, Infant Outcomes Monitoring, Research, and Prevention Branch, Division of Birth Defects and Infant Disorders, National
Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention
Internal Peer Reviewers
Tom Brody, EPA Region 5
Linda Harwell, EPA ORD, Center for Environmental Measurement and Modeling
This document has been reviewed by the U. S. Environmental Protection Agency, Office of Research and Development, and approved
for publication. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.
-------
Table of Contents
1.0 Overview of Report 1
2.0 Background 3
Brief Overview of EQI 2000-2005 3
EQI 2000-2005, Summary of Creation 3
3.0 Development of the EQI 2006-2010 5
Overview 5
Data Source Identification and Review 5
Approach 5
Summary of Activities 6
Built-Environment Domain 12
Summary of Changes to 2006-2010 data sources from original 2000-2005 EQI 13
Variable Construction 13
Approach 13
Summary of Activities 14
Changes to 2006-2010 variable construction from original 2000-2005 EQI 19
Data Reduction and Index Construction 21
Overall Approach 21
Results 23
Changes to 2006-2010 index construction from original 2000-2005 EQI 36
Domain-Specific Index Description and Loadings on Overall EQI 37
4.0 Discussion 39
Summary of changes made to 2006-2010 version compared with 2000-2005 39
Strengths and Limitations 39
Conclusion 41
5.0 References 43
Appendix I: List of References Related to 2000-2005 Environmental Quality Index A-l
Appendix II: Identified Variables by Source for Each Domain B-l
Appendix III: Changes in Variables from EQI 2000-2005 to EQI 2006-2010 C-l
Appendix IV: Table of Highly Correlated Variables for Each Domain D-l
Appendix V: Sociodemographic and Built-Domain Valence Correction E-l
Appendix VI: County Maps of Environmental Quality Index 2006- 2010 F-l
Appendix VII: Quality Assurance G-l
-------
List of Table
Table 1. Constructs for each environmental domain 5
Table 2 Sources of data for air, water, land, built-environment, and sociodemographic domains for use in the county
Environmental Quality Index 20006-2010 7
Table 3. 2005 NATA variables included in EQI 2006-2010 15
Table 4. Air domain variable means, standard deviations (SDs), and ranges - Overall and rural-urban continuum codes
(RUCCs) stratified 23
Table 5. Water domain variable means, standard deviations (SDs), and ranges - Overall and rural-urban continuum codes
(RUCCs) stratified 25
Table 6. Land domain variable means, standard deviations (SDs), and ranges - Overall and rural-urban continuum codes
(RUCCs) stratified 28
Table 7. Sociodemographic domain variable means, standard deviations (SDs), and ranges - Overall and rural-continuum
codes (RUCCs) stratified 29
Table 8. Built-environment domain variable means, standard deviations (SDs), and ranges - Overall and rural-urban
continuum codes (RUCCs) stratified 30
Table 9. Variable loadings, valence determination of variables - Air domain 31
Table 10. Variable loadings, valence determination of variables - Water domain 32
Table 11. Variable loadings, valence determination of variables - Land domain 34
Table 12. Valence corrected variable loadings, valence determination of variables - Sociodemographic domain 35
Table 13. Valence corrected variable loadings, valence determination of variables - Built domain 36
Table 14. Description of the domain indices contributing to the overall and rural-urban continuum codes (RUCCs)
stratified Environmental Quality Index for 3143 U.S. counties (2006-2010) 37
Table 15. Loadings of the domain indices contributing to the overall and rural-urban continuum codes (RUCCs)
stratified Environmental Quality Index for 3143 U.S. counties (2006-2010) 38
-------
List of Figures
Figure 1. Conceptual environmental quality - Hazardous and beneficial aspects 3
Figure 2. Principal component analysis for the Environmental Quality Index (EQI). All counties included with four
rural-urban continuum codes (RUCCs) 4
Figure 3. Rural-urban continuum code (RUCC) stratification for all counties in the United States 23
-------
-------
List
of Acronyms
ACRES Assessment, Cleanup, and Redevelopment RCRA
Exchange RqE
AQS Air Quality System RUCC
C-CAP Coastal Change Analysis Program ^
CO Carbon monoxide SDWIS
CWA Clean Water Act gQ2
EPA United States Environmental Protection Agency SSTS
EQI Environmental Quality Index TIGER
FARS Fatality Annual Reporting System
FBI UCR Federal Bureau of Investigation Uniform Crime TOD
Report -ppj
FIPS Federal Information Processing Standard
GIS Geographic information systems y g
GTFS General Transit Feed Specification USDA EE
HAP Hazardous air pollutant
HUD Housing and Urban Development WATERS
LEHD Longitudinal Employer-Household Dynamics
LQG Large Quantity Generators WQS
MRLC Multi-Resolution Land Characteristics
MSHA Mine Safety Health Administration
NADP National Atmospheric Deposition Program
NATA National-Scale Air Toxics Assessment
NCOD National Contaminant Occurrence Database
NGS National Geochemical Survey
NLCD National Land Cover Database
N02 Nitrogen dioxide
NPDES National Pollutant Discharge Elimination System
NPL National Priorities List
NPUD National Pesticide Use Database
NWI National Walkability Index
PCA Principal component analysis
PM Particulate matter
PM10 Particulate matter below 10 micrometers (|im) in
aerodynamic diameter
PM2 5 Particulate matter below 2.5 micrometers (|im) in
aerodynamic diameter
PWS Public water systems
RAD REACH Address Database
Resource Conservation and Recovery Act
Report on the Environment
Rural-urban continuum code
Standard deviation
Safe Drinking Water Information System
Sulfur dioxide
Section Seven Tracking System
Topologically Integrated Geographic Encoding and
Referencing
Transit Oriented Development
Toxic Release Inventory
Treatment, Storage, and Disposal
United States
United States Department of Agriculture Economic
Research Service
Watershed Assessment, Tracking, and
Environmental Results
Water quality standards
-------
-------
1.0
Overview of Report
An overall Environmental Quality Index (EQI), which represents
multiple domains of the ambient environment, including air,
water, land, built, and sociodemographic, for all counties in the
United States, was created for the period 2000-2005[l], It was
developed to provide a better estimate of overall environmental
quality and to improve the understanding of the relationship
between environmental conditions and human health. This report
describes the efforts to update the EQI for all counties in the
United States for the 2006-2010 period. The EQI was created for
two main purposes: (1) as an indicator of ambient conditions/
exposure in environmental health modeling and (2) as a covariate
to adjust for ambient conditions in environmental models.
However, with the public release of the EQI and variables that
constructed the EQI, other uses may emerge. The methods
applied provide a reproducible approach that capitalizes almost
exclusively on publicly available data sources.
This report is written for audiences interested in the construction
of the EQI and is technical in nature. The created variables,
EQI, domain-specific indices, and EQI stratified by rural-urban
continuum codes (RUCCs) are available publicly at the United
States Environmental Protection Agency's (EPA's) Environmental
Dataset Gateway. Also, an interactive map of the EQI is available
at EPA's GeoPlatfonii.
-------
-------
2.0
Background
Conceptually, the EQI accounts for the multiple domains of the
environment with which humans interact (see Figure 1). These
domains include chemical, natural, built, and sociodemographic
environments that have both positive and negative influences on
health. People move in and out of these positive and negative
influences. Also, the positive and negative influences often are
co-located.
Brief Overview of EQI 2000-2005
The EQI 2000-2005 was developed in four steps: (1) The five
domains were identified, (2) data for each of the five domains
were located and reviewed, (3) environmental variables were
developed from the data sources, and (4) data were combined in
each of the environmental domains; then these domain indices
were used to create the overall EQI. The EQI relied on data
sources that were mostly available to the public. Below is a
Summary of the creation of the county level EQI 2000-2005.
For more detailed technical information, see the technical report
for EQI 2000-2005 [1] located at the Environmental Dataset
Gateway.
EQI 2000-2005, Summary of Creation
Domain Identification. Based on three sources, (1) the Report
on the Enviromnent (ROE) |2J. (2) literature review, and
(3) experts, five environmental domains were identified and
developed for the EQI: (1) air, (2) water, (3) land, (4) built, and
(5) sociodemographic.
Data Source Identification and Review. Predetermined constructs
were identified to represent each domain. Based on those
constructs, data sources were explored to provide variables
representing those constructs.
Air Domain: Three data types were considered: (1) monitoring
data, (2) emissions data, and (3) modeled estimates representing
two constructs: concentrations of either criteria air pollutants
or hazardous air pollutants (toxics). Twelve data sources were
identified, and seven were considered for the EQI. Two were
used for the air domain of the EQI because they were the most
complete.
Water Domain: Five broad data types witliin the water domain
were identified: (1) modeled, (2) monitoring, (3) reported, (4)
survey/study, and (5) miscellaneous data. Eighty data sources
were identified. Five were used for the water domain of the
EQI representing seven constructs: water quality, general
water contamination recreational water quality, domestic use,
deposition, drought, and chemical contamination.
Land Domain: Land domain data sources were grouped into five
constructs: (1) agriculture, (2) pesticides, (3) contaminants, (4)
facilities, and (5) radon. Eighty sources were identified. Eleven
were retained.
ENVIRONMENTAL QUALITY
Hazardous Beneficial
Polluted Air
Home Ownersl
Factories
'hysicai Activity
Figure 1. Conceptual environmental quality - Hazardous and
beneficial aspects.
Built-Environment Domain: Built environment considered five
data types: (1) traffic-related. (2) transit access, (3) pedestrian
safety, (4) access to various business environments (such as the
food, recreation, health care, and educational environments),
and (5) the presence of subsidized housing. Twelve data sources
were identified, and four were retained for the built-enviromnent
domain of the EQI for five constructs: (1) roads, (2) highway
road safety, (3) public transit behavior, (4) business environments
(physical activity, food, health care, and educational), and (5)
subsidized housing.
Sociodemographic Domain: The sociodemographic domain is
represented by crime and socioeconomic constructs. Only two
data sources were identified for the sociodemographic domain of
the EQI, one for each of the constructs.
Variable Construction. After researching and choosing data
sources, variables were created to represent each of the five
domains. New variables were created because raw data sources
were not always appropriate for statistical analysis.
The process for selecting and creating variables included
making variables for each domain for each available year of
data (2000-2005),
looking for highly correlated variables that are giving the same
information statistically and deciding which of the variables
best represents the environmental domain (and remove the
extra variables).
3
-------
Principal components
analysis (PCA) reduced
multiple variables into
domain-specific indices
for each RUCC strata and
overall.
Domain-specific indices
combined using PCA to
create EQI for each
RUCC strata and overall.
Air
variables
Water
variables
Socio-
demographic
variables
Land
*
variables
Built
variables
EQI
Air
Indices
Built
Indices
Socio-
demographic
Indices
Water
Indices
Land
Indices
RUCC1 = metropolitan-urbanized
RUCC2 =nonmetropolitan-urbanized
RUCC3 =less urbanized
RUCC4 =thinly populated
OVERALL
Figure 2. Principal component analysis for the Environmental Quality Index (EQI). All counties included with four rural-urban
continuum codes (RUCCs).
looking for missing data,
looking at the distribution and statistical properties of each
variable and deciding how it should be scaled for analysis, and
averaging variables from 2000-2005 for each county.
Data Reduction and Index Construction. After variables were
created, they were combined into a single index (the EQI) using
statistical methods. Each domain has its own index (air domain
index, water domain index, etc.). Next, each of the domain-
specific indices was used to create the overall EQI. The statistical
process used to add these variables together is called principal
component analysis (PCA). Figure 2 shows the steps that include
Since the creation of the EQI 2000-2005, multiple studies
were conducted examining the relationship between overall
environmental quality and health outcomes, including preterm
birth [3], mortality |4|. cancer incidence |5|. asthma prevalence
[6], physical inactivity and obesity [7], infant mortality [8], and
pediatric multiple sclerosis [9], A complete list of references
related to EQI and health outcomes is shown in Appendix I.
4
-------
3.0
Development of the EQI 2006-2010
Overview
The development of the EQI 2006-2010 followed mostly the
same protocol as the EQI 2000-2005. The majority of constructs
identified for each of the five domains in the EQI 2000-2005
were maintained as the basis for variable identification, with
the exception of one deletion each in the water domain and land
domain and constructs added to the water domain land domain
sociodemographic domain, and the built-enviromnent domain.
Most data sources remained unchanged. Principal components
analysis was used to develop the indices. However, using
lessons learned from the creation of the EQI 2000-2005, some
modifications were adopted to improve the EQI 2006-2010; these
modifications included exploring new data sources that were not
available during EQI 2000-2005 development, assessment of
all variables for continued inclusion in the EQI, and assessment
of variables' valence within a domain and valence correction.
This section outlines the development of the EQI 2006-2010
through (1) data source identification and review, (2) variable
construction, and (3) data reduction and index construction.
Data Source Identification and Review
Approach
Data Selection
An index that comprehensively captures the total enviromnent
relating to human health requires numerous variables
representing the full range of health-influencing exposures.
From within each domain identified in the conceptual model (air,
water, land, sociodemographic, and built enviromnents), specific
constructs or major areas were identified (Table 1). In general, the
identified constructs from EQI 2000-2005 were maintained for
the EQI 2006-2010. However, in the water domain, we removed
the "recreational water quality" construct as it only provided
data for 231 counties in the United States with beach recreational
waters. Because of this low representation, the variables in
this domain had extremely low loading values in the Principal
Components Analysis; therefore, they were removed in the
2006-2010 EQI. In addition, a dataset representing drinking water
quality was identified and, therefore, we were able to include
"Drinking water quality" construct. In the land domain, the
"Contaminants" construct was eliminated. We eliminated these
data because they were not the same quality as the rest of the
data for the EQI. There was a lack of updated contaminants data.
Table 1. Constructs for each environmental domain.
Domain
Air
Water
Land
Sociodemographic
Built Environment
Constructs
Criteria air pollutants
Hazardous air pollutants
Overall water quality
General water contamination
Domestic use
Atmospheric deposition
Drought
Chemical contamination
Drinking water quality (new 2006-2010)
Agriculture
Pesticides
Facilities
Radon
Mining activity (new 2006-2010)
Socioeconomic
Crime
Political character (new 2006-2010)
Creative class representation (new 2006-2010)
Roads
Highway/road safety
Commuting behavior
Business environment
Housing environment
Walkability (new 2006-2010)
Green space (new 2006-2010)
5
-------
and, because of the high correlation between this construct and
constructs in other domains, contaminants of this type were better
represented by water contaminant data. Also, in the land domain,
a "Mining activity" construct was added. The sociodemographic
domain added two new constructs: (1) political character and
(2) creative class representation. There was a change in how
educational attainment was represented in the 2006-2010 EQI.
That change in the education variable from percent of adults
with greater than high school education in the 2000-2005 EQI
to percent of adults with a college education in the 2006-2010
EQI resulted from inclusion of an education variable with more
variability, as almost all citizens have a high school education
at this time. The built-environment domain added two new
constructs: (1) walkability and (2) green space. Data sources
were explored to identify variables that represent the identified
constructs for construction. All data sources used for EQI 2000-
2005 were reviewed for data updates, and a subsequent search
was conducted to identify potential new data sources.
We had solid representation of data for most domains, and we
sought to ensure continuity and comparability for the 2006-2010
EQI. Still, our update required identification of new data sources
to ensure representation of identified constructs. Because the
team came to appreciate the limitations and knowledge gaps in
data from the original EQI, the data source identification process
was different for the 2006-2010 period than that undertaken
for the original (2000-2005) EQI. For example, because of
limitations in the National Geochemical Survey representing the
geology construct in the land domain, we looked for alternative
sources and are now using mines data in the land domain. In
recognition of gaps, such as the absence of walkability in the built
domain and absence of political climate in the sociodemographic
domain, we sought additional data sources to represent the new
constructs that we believed would represent more fully the
environmental quality of a county.
The details of the new data sources that were identified and
included in the EQI 2006-2010 are included in the data source
descriptions below.
Data Source Search
Once the desired constructs were identified, the research team
conducted an extensive search for potential sources for data
to represent those constructs. In general, a broad approach to
searching for data sources was undertaken to
identify EPA and non-EPA domain-specific environmental data
sources for all counties in the 50 states of the United States;
summarize environmental data source availability, quality,
spatial and temporal coverage, storage requirements, and
acquisition steps; and
Possible data sources were identified using Web-based search
engines (e.g., Google), site-specific search engines (e.g.,
federal and state data sites), literature-reported data sources
(e.g., PubMed, ScienceDirect, TOXNET), and personal
communications from data owners. Data that were available at or
had the potential to be aggregated to the U.S. county level were
sought. Data were restricted to represent the years 2006-2010.
Data Quality and Coverage Assessment
Once potential data sources were identified, several criteria were
used to assess sources for inclusion in the EQI. First, constructs
representing each domain were identified. Data sources were
evaluated as to whether variables could be developed to represent
the construct. If a data source could provide variables for a
construct in the domain, then data quality and data coverage were
used to evaluate data sources for use in the EQI. Data sources
of the highest quality were sought. Quality was assessed by one
or more of the following ways: Through documentation and
discussion with the data source managers, in data reports and
internal documentation, project investigators, and the larger field
of environmental research through use and critique of the various
data sources. Data coverage, which included spatial and temporal
components, was more challenging to achieve. Coverage
for the entire United States, including Alaska and Hawaii,
was one important spatial criterion. Often, it was relatively
straightforward to identify high-quality data on a few individual
locations or a small geographic area, but the EQI was developed
to represent all counties (N=3143) in all 50 States. A second
spatial criterion was county-level representation, so data had to
be constructible at the county-level for inclusion (e.g., average
of point measures or census tract values). Temporally, ideal
sources would have had annual data for the 2006-2010 period. At
minimum, however, at least some data must have fallen within
the 2006-2010 period or close to this time. In theory, a "perfect"
data source would have variable measurements at high temporal
and spatial resolutions. In practice, data often met one but not
both criteria, and evaluation of trade-off values was required,
along with consideration of data quality. Unfortunately, some of
the data sources used in EQI 2000-2005 did not have any updates
for the 2006-2010 period. Redundant data sources that were
determined to meet the criteria for inclusion but were not selected
for inclusion were retained for use in sensitivity analyses.
Summary of Activities
Table 2 identifies the data sources that were acquired and used for
the construction of the EQI and includes a description of the data
source and variables constructed from data source.
obtain the identified data.
6
-------
Table 2 Sources of data for air, water, land, built-environment, and sociodemographic domains for use in the county Environmental
Quality Index 20006-2010
Air Domain
Source of Data
Air Quality System
(AOS 2006-2010)
[10]
Description
Repository of ambient air quality data, including both
criteria and hazardous air pollutants (HAPs)
Variables*
PM|0 - Particulate matter under 10 |jg in aerodynamic
diameter (|jg/mJ 5-year average); PM,5 - Particulate
matter under 2.5 |jg in aerodynamic diameter (|jg/
md 5-year average); NO,- Nitrogen dioxide (parts per
billion [ppb] 5-year average); SO, - Sulfur dioxide (ppb
5-year average); 03 - Ozone (parts per million (ppm)
5-year average); CO - Carbon -monoxide (ppm 5-year
average)
EQI version
2000-2005 and updated 2006-2010
National-Scale Air
Toxics Assessment
(NATA 2005)[11]
Estimates of HAP concentrations using emissions
information from the National Emissions Inventory and
meteorological data input into the Assessment System for
Population Exposure Nationwide model
Water Domain
Source of Data
Watershed
Assessment,
Tracking and
Environmental
Results Program
Database
(WATERS)[12]
A_TeCA -1,1,2,2-tetrachloroethane (tons emitted per
year); A_112TCA-1,1,2-trichloroethane (tons emitted
per year); A_DBCP -1,2-dibromo-3-chloropropane
(tons emitted per year); A_Acrylic_acid-Acrylic acid
(tons emitted per year); A_Benzidine - Benzidine (tons
2000-2005 and 2006-2010 (used 2005
NATA only)
emitted per year)
emitted per year)
emitted per year)
Description
Collection of EPA water assessments programs, including
impairment, water quality standards, pollutant discharge
permits, and beach violations
; A_Benzyl_CI - Benzyl chloride (tons
; A_Be - Beryllium compounds (tons
; A_DEHP - b/s-2-ethylhexyl phthalate
(tons emitted per year); A_CCI4 - Carbon tetrachloride
(tons emitted per year); A_CS - Carbon sulfide (tons
emitted per year); A_CI - Chlorine; A_C6H5CI -
Chlorobenzene (tons emitted peryear); A_chloroform
- Chloroform (tons emitted per year); A_Chloroprene
- Chloroprene (tons emitted per year); A_Cr - Chromium
compounds (tons emitted peryear); A_Co - Cobalt
compounds (tons emitted peryear); A_CN - Cyanide
compounds (tons emitted peryear); A_DBP -
Dibutylphthalate (tons emitted per year); A_EtCI - Ethyl
chloride (tons emitted per year); A_EDB - Ethylene
dibromide (tons emitted per year); A_EDC - Ethylene
dichloride (tons emitted per year); A_Formaldehyde -
Formaldehyde (tons emitted per year); A_Glycol_ethers
- Glycol ethers (tons emitted per year); A_N2H2 -
Hydrazine (tons emitted per year); A_HCI - Hydrochloric
acid (tons emitted per year); AJsophorone - Isophorone
(tons emitted per year); A_Mn - Manganese compounds
(tons emitted peryear); A_MeBr - Methyl bromide
(tons emitted peryear); A_MeCI - Methyl chloride (tons
emitted per year); A_PH3 - Phosphine (tons emitted per
year); A_PCBs - Polychlorinated biphenyls (tons emitted
per year); A_ProCI2 - Propylene dichloride (tons emitted
per year); A_Quinolin - Quinoline (tons emitted per
year); A_C2HCI3 - Trichloroethylene (tons emitted per
year); A_VyCI - Vinyl chloride (tons emitted peryear)
Variables!
ALLNPDESperKMJn - All NPDES permits per 1000 km
of stream in county (permits per 1000 km stream length)
EQI version
2000-2005 and updated 2006-2010
7
-------
Table 2. continued
National
Atmospheric
Deposition Program
(NADP 2006-2010)
[13]
Estimated Use of
Water in the United
States (2010)[14]
Drought Monitor
Data (2006-2010)
[15]
Samples both regulated and unregulated contaminants
in public water supplies; maintained by EPA to satisfy
statutory requirements for Safe Drinking Water Act
County-level estimates of water withdrawals for domestic,
agricultural, and industrial use calculated by the United
States Geological Survey
Geographic information systems raster files reporting
weekly modeled drought conditions; a collaboration
that includes the National Atmospheric and Oceanic
Administration, the U.S. Department of Agriculture, and
academic partners.
Measures deposition ofvarious pollutants, such as
calcium, sodium, potassium, and sulfate, from rainfall
Safe Drinking Water Monitoring of public water systems for health-based
Information System violations
(SDWIS 2006-2010)
[17]{United States
Environmental
Protection Agency
(EPA), #966}
Land Domain
Source of Data Description
National Pesticide Delineates state-level pesticide usage rates for cropland
Use Database: applications; contains estimates for active ingredients, of
2009[18] which 68 are insecticides, and 22 are other pesticides
CaAveJn - Calcium (Ca) precipitation weighted mean
(mg/L); KAveJn - Potassium (K) precipitation weighted
mean (mg/L); N03Ave - Nitrate (NO-J precipitation
weighted mean (mg/L); ClAveJn - Chloride (CI)
precipitation weighted mean (mg/L); S04_mean_ave
- Sulfate (S04) precipitation weighted mean (mg/L);
HgAve - Total mercury deposition (ng/M J
Per_TotPopSS - Percent of population on self supply
(percent); Per_PSWithSW - Percent of public supply
population that is on surface water (percent)
2000-2005 and updated 2006-2010
AvgOfD3_ave - Percent of county drought -
(D3-D4) (percent)
W_As_ln -Arsenic (mg/L); W_Ba_ln - Barium (mg/L);
W_Cd_ln - Cadmium (mg/L); W_Cr_ln - Chromium
(total) (mg/L); W_CN_ln - Cyanide (mg/L); W_FL_ln -
Fluoride (mg/L); W_HG_ln - Mercury (inorganic) (mg/L);
W_N03_ln - Nitrate (as N) (mg/L); W_N02_ln - Nitrite
(as N) (mg/L); W_SE_ln - Selenium (mg/L); W_Sb_ln
-Antimony (mg/L); W_Endrin_ln - Endrin (|jg /L); W_
methoxychlorjn - Methoxychlor (ug/L); W_Dalapon_ln -
Dalapon (|jg /L); W_DEHA_ln - Di(2-ethylhexyl)adipate
(DEHA) (|jg /L); W_Simazine_ln - Simazine (|jg /L);
W_DEHP_ln - Di(2-ethylhexyl) phthalate (DEHP)( pg
/L); W_Picloram_ln - Picloram (|jg /L); W_Dinoseb_ln
- Dinoseb (|jg /L); W_atrazine_ln - Atrazine (|jg
/L); W_24D_ln - 2,4-D (2,4-Dichlorophenoxyacetic
acid) (|jg /L); W_BenzoAP_ln - Benzo[a]pyrene
(|jg /L); W_PCP_ln - Pentachlorophenol (|jg /L);
W_PCB_ln - Polychlorinated biphenyls (PCBs) (|jg
/L); W_DBCP_ln -1,2-Dibromo-3-chloropropane
(DBCP) (|jg /L); W_EDB_ln - Ethylene dibromide
(EDB) (|jg /L); W_xylenes_ln - Xylenes (Total)( |jg
/L); W_Chlordane_ln - Chlordane (|jg /L); W_DCM_ln
- Dichloromethane (methylene chloride) (|jg /L); W_
PDCBJn -1,4-Dichlorobenzene (p-dichlorobenzene)
(|jg /L); W_111 trichlorane_ln -1,1,1 -Trichloroethane
(|jg /L); W_Trichlorene_ln - Trichloroethylene (|jg
/L); W_C2CI4_ln - Tetrachloroethylene (|jg /L); W_
benzenejn - Monochlorobenzene (chlorobenzene) (|jg
/L); W_Toluene_ln - Toluene (|jg /L); W_ethylbenz_ln
- Ethylbenzene (|jg /L); W_styrene_ln - Styrene (|jg
/L); W_Alpha - Alpha Particles (Gross Alpha, excluding
radon and uranium) (pCi/L); W_DCE_ln - cis-1,2-
Dichloroethylene (|jg /L)
Coliform_proportion_ln - Total coliform proportion
(average number of violations*(population served/
county population)
Variablest
insecticidejn - Insecticide applied (lb); herbicidejn-
2000-2005 and updated 2006-2010
extreme 2000-2005 and updated 2006-2010
2000-2005 and 2006-2010 (not
updated, used same variables from
2000-2005)
2006-2010
EQI version
2000-2005 and updated 2006-2010
Herbicides applied (
applied (lb)
; fungicidejn - Fungicides
8
-------
Table 2. continued
2007 Census of
Agriculture Full
Report[19]
Summary of agricultural activity including number of
farms by size and type, inventory and values for crops and
livestock, and operator characteristics
pct_manure_acres_ln - Manure, acres applied per
county acres (percent); pct_nematode_acres_ln -
Chemicals used to control nematodes, acres applied
per county acres (percent); pct_disease_acres_ln
- Chemicals used to control diseases in crops and
orchards, acres applied per county acres (percent);
pct_defoliate_acres_ln - Chemicals used to control
growth, thin fruit, or defoliate, acres applied per county
acres (percent); Pct_AU_ln - Animal units, animal
units per county acres (percent); farms_per_acre_ln
- Number of farms (number); pct_irrigated_acres_ln
- Irrigated acres, acres irrigated per county acres
(percent); pct_harvested_acres_ln - Harvested acres,
acres harvested per county acres (percent)
2000-2005 and updated 2006-2010
EPA Geospatial Data
Download Service
(2006-2010)[20]
Maintained by EPA and provides locations of and
information on facilities throughout the United States;
different datasets within this database are updated at
different intervals, but most are updated monthly; no set
spatial scale across datasets. Some provide addresses,
some geocoded addresses, etc.
facilities_rate_ln - Log transformed rate of all facilities
per county (proportion)
2000-2005 and updated 2006-2010
Map of Radon
Zones [21]
Identifies areas of the United States with the potential for
elevated indoor radon levels; maintained by EPA
Radon - Radon zone (ordinal value)
2000-2005 and 2006-2010 (not
updated, used same variable from
2000-2005)
Mine Safety
and Health
Administration
(MSHA) Mines Data
Set(2006-2010)[22]
Includes status of coal/metal/nonmetal mines under MSHA
jurisdiction since 1970
std_coal_prim_pop_ln - Primarily coal mines, mines per
county population (proportion); std_metal_prim_pop_ln
- Primarily metal mines, mines per county population
(proportion); std_nonmetal_prim_pop_ln - Primarily
nonmetal mines, mines per county population
(proportion); std_sandandgravel_prim_pop_ln -
Primarily sand and gravel mines, mines per county
(proportion); std_stone_prim_pop_ln - Primarily stone
mines, mines per county population (proportion)
2006-2010
National
Geochemical
Survey[23]
Geochemical data (arsenic, selenium, mercury, lead, zinc,
magnesium, manganese, iron, etc.) for the United States
based on stream sediment samples
2000-2005; not used in 2006-2010.
These data are represented in the
water domain with the National
Contaminant Occurrence Database
(2006-2010) and the National
Atmospheric Deposition Program
(2006-2010)
Sociodemographic Domain
Source of Data
Description
Variables^
EQI version
United States
Census (2010)[24]
County-level population and housing characteristics,
including density, race, spatial distribution, education,
socioeconomics, home and neighborhood features, and
land use
Pct_RenterOcc - Percent renter-occupied units
(percent); Pct_Vacant_Housing - Percent vacant units
(percent); Med_HH_Value - Median household value
(dollars); ln_HH_lnc - Natural log transformed median
household income (dollars); pct_fam_pov - Percent
of families living below federal poverty level (percent);
pct_BS - Percent of persons with bachelor's degree or
higher, age 25+ (percent); pct_unemp_total - Percent of
persons who are unemployed (percent); ln_Occs_Room
- Natural log transformed number of occupants per
room (count); GINI_est - Measure of income inequality
(proportion)
2000-2005 and updated 2006-2010
Uniform Crime
Reports (2006-2010)
[25]
County-level reports of violent crime
ln_ViolAv - Natural log transformed violent crime rate
(log of count ofviolent crimes / county population)
2000-2005 and updated 2006-2010
Dave Leip's Atlas
of U.S. Presidential
Elections (2008)[26]
2008 Election results
DEM02008 - Percent county voting Democratic in 2008
(percent)
2006-2010
9
-------
Table 2. continued
United States An index of a county's share of population employed in
num_CreatClass - Percent county employed in a
2006-2010
Department occupations that require thinking creatively"
creative class (percent)
of Agriculture
Economic Research
Service Creative
Class County Codes
(2010)|27]
Built-Environment Domain
Source of Data
Dun and
Bradstreet North
American Industry
Classification
System codes
(2008)[28]
Description
Description of physical activity environment (recreation
facilities, parks, physical-fitness-related businesses), food
environment (fast food restaurants, groceries, convenience
stores), and education environment (schools, daycares,
universities) per county
Topologically
Integrated
Geographic
Encoding and
Referencing (2009)
[29] and NAVTEQ
map data[30]
Fatality Annual
Reporting System
(2006-2010) [31 ]
Housing and Urban
Development Data
(2010)[32]
Road type and length per county; road types by county
created by joining NAVTEQ map data to Topologically
Integrated Geographic Encoding and Referencing (TIGER)
county definitions
Annual pedestrian-related fatality per 100,000 population;
maintained by National Highway Safety Commission
Housing authority profiles provide general housing details
(low-rent and subsidized/Section 8 housing); information
updated by individual public housing agencies.
Variables!
al_pwn_gm_env_rate_ln - Natural log transformed rate
of vice-related businesses per county (log of count
of businesses / county population); ed_env_rate_ln
- Natural log transformed rate of education-related
businesses per county (log of count of businesses /
county population); neg_food_rate_ln - Natural log
transformed rate of negative food resources per county
(log of count of businesses / county population); pos_
food_rate_ln - Natural log transformed rate of positive
food resources per county (log of count of businesses
/ county population); hc_env_rate_ln - Natural log
transformed rate of health-care-related businesses per
county (log of count of businesses / county population);
rec_env_rate_ln - Natural log transformed rate of
recreation-related businesses per county (log of count
of businesses / county population); ss_env_rate_ln -
Natural log transformed rate of social service agencies
per county (log of count of businesses / county
population); civic_env_rate_ln - Natural log transformed
rate of civic-related businesses per county (log of count
of businesses / county population
SecondaryRoadProportion - Proportion of all roads that
are secondary roads (proportion)
EQI version
2000-2005 and updated 2006-2010
Ln_fatalities - Natural log transformed rate (count/
county population) of fatal car crashes per county (log
transformed count / county population)
total_units_ln - Natural log transformed rate of the sum
of the following two variables (Iow_rent_un its - Count
of low-rent units per county [count] and section_eight_
units - Count of section eight units per county [count])
(log of summation of units / county population)
2000-2005 and updated 2006-2010
2000-2005 and updated 2006-2010
2000-2005 and updated 2006-2010
United States
Census (2010)[24]
EnviroAtlas Green
space dataset (2011,
2005-2011)[33]
EPA's National
Walkability Index
) (2010)[36]
County-level population characteristics, including density,
race, spatial distribution, education, socioeconomics, home
and neighborhood features, and land use
Description of 20 different land covers for National Land
Cover Database (NLCD)[34] and 24 for Coastal Change
Analysis Program (C-CAP)[35]; given as percent of county
Characterizes every census block group walkability on
a score from 0 to 20 based on four variables: (1) mix
of employment types and occupied housing, (2) mix of
employment types in a block group, (3) street intersection
density, and (4) predicted commute mode split - proportion
of workers in the block group who carpool
CommuteTime - Time it takes to travel from home to 2006-2010
work (min); ln_PubTrans -Natural log of percent of
county residents who report using public transportation
(percent)
NINDEX_open - Percent of county land area classified 2006-2010
as natural land cover and open space developed land
cover (percent)
sum_NWIBG - Walkability score (ordinal) 2006-2010
"Air domain: All variables are natural log transformed with the exceptions of A_edb,
A_formaldehyde, 03, PM|0, and PM-,5.
Water, Land, and Built domains: Variables with Jn indicated natural log transformation.
Sociodemographic domain: ln_ indicates natural log transformation.
Data sources highlighted in blue are new data sources added to 2006-2010 EQI. Data
sources highlighted in gree are data sources used in 2000-2005 EQI but are not included
in 2006-2010 EQI.
10
-------
Air Domain
Two constructs represent the air domain: (1) criteria air
pollutants and (2) hazardous air pollutants (HAPs). The Air
Quality System (AQS)[10] was used to construct variables for
the criteria air pollutants and the National-Scale Air Toxics
Assessment (NATA) database [11] was used to construct
variables for the HAPs.
The AQS is a repository for criteria ambient air pollution
data collected by federal, state, local, and tribal agencies
from thousands of monitors for EPA's ambient air monitoring
program across the United States. Monitored pollutants include
all criteria air pollutants, PM species, and approximately 60
ozone precursors. Major strengths of the AQS are that data
are measured, rather than modeled, and these measurements
are synchronized across the country. Monitors in the network
and the reported data are audited regularly for accuracy and
precision. However, most of the ambient air monitors are located
in or near urban areas, leaving many U.S. counties without
reported data. In addition, the AQS provides sparse and limited
data collection for HAPs.
The NATA database uses data from the National Emissions
Inventory [3 7] to construct air dispersion models for estimating
ambient concentrations of HAPs at the county and census-tract
levels. Beginning in 1996, the National Emissions Inventory
data are constructed every 3 years, providing annual estimates.
The NATA databases contain estimated ambient concentrations
for 177 to 180 of the 187 HAPs and use validated models that
take meteorology and chemical dispersion into account. The
methodology for estimating concentrations may change between
assessments, but these modifications are well documented
and justified. Although the ambient concentrations may be
comparable over time, some differences between estimates
are attributable to these minor methodological modifications.
The temporal resolution of the assessments is adequate for the
intended EQI, but, because of the 3-year release schedule, there
are gaps in temporal coverage. NATA 2008 was not developed
and thus, for EQI 2006-2010, NATA 2005 was used.
Water Domain
The water domain included six data sources: (1) the WATERS
program database[12], (2) Estimated Use of Water in the United
States. [14], (3) the National Atmospheric Deposition Program
(NADP)[13], (4) the Drought Monitor Network[ 15], (5) the
National Contaminant Occurrence Database (NCOD)[16], and
(6) the Safe Drinking Water Information System (SDWIS)[17],
Using these six data sources, variables were created to represent
seven constructs that describe the overall water environment.
The seven constructs were (1) overall water quality, (2) general
water contamination, (3) drinking water quality, (4) domestic
use, (5) atmospheric deposition, (6) drought, and (7) chemical
contamination.
The Watershed Assessment, Tracking, and Environmental
Results (WATERS) Program[12] database represents the surface
water assessment programs under the Clean Water Act (CWA).
A limitation of this data source is that data are maintained at
the state level and reported to the federal system. Although
all states report county-level data, there is little consistency in
the temporal reporting and type of data reported across states.
These data were first geocoded to a specific stream length in
the National Hydrography Dataset[38] via the REACH Address
Database (RAD) [3 9], The geocoded WATERS program data
were used to calculate human-exposure-related variables, such
as percentage of stream length impaired for recreational use.
This dataset is the only database maintaining information on EPA
CWA regulations, which is a strength.
The National Contaminant Occurrence Database (NCOD)[16] is
a surveillance database maintained to satisfy the requirements of
the Safe Drinking Water Act. This database includes information
on contaminants in public water supplies that are not measured
elsewhere. The survey is conducted every 6 years, and data are
provided by public water suppliers. The data are limited, as they
are provided by public water suppliers, and, therefore, spatial
aggregation was needed to get county-level estimates. Estimated
Use of Water in the United Stales\ 14|. which is modeled by the
United States Geological Survey, provided county-level estimates
of water withdrawals (an indication of water stress in a county)
for domestic, irrigation, livestock, and industrial use. This dataset
already is provided at the county level, which is a strength;
however, it is limited, as the estimates are based on several
different data sources.
Two data sources provided information on meteorological
impacts on water quality. The Drought Monitor Data[15] are
modeled weekly drought conditions. Weekly coverage for
the entire country is a strength of this dataset. The National
Atmospheric Deposition Program (NADP)[13] provided weekly
measures and national coverage of the deposition of various
pollutants from rainfall using monitors around the country. Again,
this database provided weekly information for the entire country;
however, it was reported by monitors and required spatial
aggregation to achieve county-level estimates.
Drinking water quality data was gathered from the Safe Drinking
Water Information System[17] (SDWIS), which is a repository
maintained for compliance with federal regulations. This is a
new data source to the water domain. SDWIS provides publicly
available data based on requirements from the Safe Drinking
Water Act. States are required to report basic information about
the public water systems (PWS), violations, and enforcement
information. The health-based violations provided in SDWIS
are not measured elsewhere. Of the SDWIS measures, only total
coliform health-based violations were considered for inclusion
in the 2006-2010 EQI, as the other contaminant categories have
a high frequency of missing data (arsenic: 87.18%; ground
water: 97.8%; inorganic chemicals: 97.04%; lead and copper:
90.87%; long-term enhanced surface water treatment rule
1 and 2: 87.69%; nitrates: 91.92%; radionuclides: 89.76%;
disinfection and disinfectant by-products: 66.43%; surface water
treatment: 90.84%; synthetic organics: 98.79%; and volatile
organic chemicals: 98.5%) for health-based violations. Average
total coliform health-based violations were used to estimate
the proportion of the county population affected by coliform
violations between 2006 and 2010.
Land Domain
The land domain included five data sources representing five
constructs: (1) Agriculture, (2) Pesticides, (3) Facilities, (4)
Radon, and (5) Mining Activity. The data sources identified
for this domain include: 2007 Census of Agriculture [19], 2009
National Pesticide Use Database[18], EPA Geospatial Data
11
-------
Download Service[20], Map of Radon Zones[21], and Mine
Safety and Health Administration (MSHA) mines data[22]. The
MSHA mines database is a data source new to EQI2006-2010.
Also, the National Geochemical Survey database used in EQI
2000-2005 was not used in EQI 2006-2010.
The 2007 Census of Agriculture Full Report[19] was used to
represent agricultural factors. Information on nonpesticide
chemicals used in farming, animal units, harvested acreage,
irrigated acreage, manure acreage, and proportion of farms
was taken from the 2007 Census of Agriculture. The Census of
Agriculture[19] data provided mostly farm-related summary
characteristics and did not offer direct pesticide measures or
probable exposure information. As a strictly environmental
indicator, the Census of Agriculture was useful, but its ability to
link to human health was somewhat limited. Eight variables from
the census of agriculture were included in the EQI.
The 2009 National Pesticide Use Database (NPUD)[18] provides
county-level rates of pesticide use. A limitation of the NPUD
was its availability only for contiguous states. Pesticides were
classified into three pesticide classes and then summed to
estimate county-level pesticide use (in kilograms) for herbicides,
fungicides, and insecticides. These three pesticide categories
were included in the EQI.
The industrial facilities data source, the EPA Geospatial Data
Download Service[21], was used to find the following types of
sites: Brownfield sites; Superfund sites; Toxic Release Inventory
sites; pesticide-producing-location sites; large-quantity generator
sites; and treatment, storage, and disposal sites. All facilities-
related data were retained for inclusion in the EQI with extensive
information on each facility for the years 2006-2010.
The EPA Radon Zone [21] map assigned a radon potential level
to each county in the United States. As the data source provided
radon potential, not actual measurement, these data were limited.
The three-level radon categorization masked important radon-
level heterogeneity across the United States. Despite these
limitations, the data sources provided land-related data not
available elsewhere.
The Mine Safety and Health Administration (MSHA) Mines
Data Set[22] was used to create the mining activity construct.
The MSHA's dataset includes current and historical coal, metal,
and nonmetal mines. The list included the status of each mine
(Abandoned, Abandoned and Sealed, Active, Intermittent, New
Mine, Nonproducing, Temporarily Idled) and in which county the
mine was located. The dataset does not include the size of each
mine, so it is possible a mine may span two counties, but only the
county indicated by its official address is reported.
The National Geochemical Survey (NGS)[23], used in the 2000-
2005 version of the EQI to determine the contaminant construct,
was not included in the updated version. The NGS data provided
the mean and standard deviations for multiple soil chemicals.
However, these values were calculated from multiple surveys of
soil samples collected over several years based on local agencies'
interests and resources and, therefore, were combining many
varying sources of data. Because of high correlation between the
NGS and the National Contaminant Occurrence Database and the
National Atmospheric Deposition Program, the decision to drop
the NGS was made.
Sociodemographic Domain
The original sociodemographic domain included only two
constructs: (1) socioeconomics and (2) crime. In an effort to
better reflect each county's sociodemographic character, the
updated Sociodemographic Domain for EQI 2006-2010 has four
constructs: (1) Socioeconomic, (2) Crime, (3) County creative
typology (new for EQI 2006-2010), and (4) County political
valence (new for EQI 2006-2010). Because counties can be
characterized as "working class" or "tech savvy," we added the
creative typology to help capture these characteristics. Similarly,
counties may be known for their political valence (e.g., a "red"
county in a "blue" state); the percent voting Democratic in the
2008 election was added to capture this county characteristic.
Only four data sources were identified and retained for the
sociodemographic domain: (1) the United States Census
Bureau[24], (2) the Federal Bureau of Investigation Uniform
Crime Reports (FBI UCRs)[25], (3) the United States Department
of Agriculture Economic Research Service (USD A ERS)[27], and
(4) Dave Leip's Atlas of U.S. Presidential Elections (2008)[26],
The United States Census[24] reports county-level population
and housing characteristics, including population density, race,
spatial distribution, socioeconomic characteristics, home and
neighborhood features, and land use. One strength of this data
source is its national coverage and consistency of data collection
with standard methods. One weakness of this data source is its
decennial collection.
The FBI UCR[25] provides annual violent and property crime
counts and rates for reporting areas. These data are a valuable
source of crime exposure, but reporting is not mandatory and may
vary by jurisdiction.
The USDAERS[27] creates a "creative class" index, derived
from census data, to identify what proportion of the population
may be employed in creative pursuits. This variable helps to
characterize counties as being attractive to people in creative
work (e.g., physicians, professors, architects). Because this
variable is based on census data, it has the same strengths and
weaknesses of the United States Census.
Dave Leip's Atlas of U.S. Presidential Elections[26] tracks the
political valence of the counties. Political valence tracks with
a number of county-level attributes, such as provision of social
supports, levels of school funding, etc. Capturing this variability
may be useful for differentiating counties from each other. One
strength of Dave Leip's Atlas of U.S. Presidential Elections data
source is its data quality, and one weakness of this data source is
its infrequency of publication.
Each of these data sources represents critical aspects of the
human sociodemographic environment and is updated regularly
and available at the county-level for the entire country.
Built-Environment Domain
Built-environment data sources were identified for the following
constructs: Business environment, Highway safety, Housing,
Roads, Commuting practices, Walkability, and Green Space. For
EQI 2006-2010, we added two new data constructs with new data
sources: one representing green space and another estimating
county walkability.
12
-------
For the road construct, NAVTEQ road map data[30] were
joined to Topologically Integrated Geographic Encoding
and Referencing (TIGER) [29] county definitions to result in
road types by county. The road data from NAVTEQ, whose
underlying map database was based on first-hand observation of
geographic features, rather than relying on official government
maps, is the majority supplier for car navigation systems
(around 85% of car makers). The TIGER files provide relatively
uniform and nationwide coverage. From these files, county-
specific proportions were characterized for various road types.
Unfortunately, considerable heterogeneity may be lost; for
instance, a tertiary road in Maryland may not be qualitatively
equivalent to one located in Wyoming.
The Fatality Analysis Reporting System[31] of the National
Highway Safety Commission was retained as part of traffic safety
because of its national coverage. The data are regularly updated
and available from the Web site. A limitation of these data is
that traffic fatalities result from diverse types of events (e.g.,
from road conditions or substance-involved fatalities), but this
diversity is not captured well.
North American Industry Classification System codes through
Dun and Bradstreet[28] were used as the data source to
estimate five different business environment topics: (1) physical
activity, (2) food, (3) educational, (4) social, and (5) health care
environments. These data are available as geocoded business
addresses. Although these data have sometimes been criticized
for inadequate spatial resolution (e.g., inaccurate geocoding to
small units of aggregation, such as census tracts), they should be
sufficient as a construct for county-level business environments
of food, physical activity, and education.
The Housing and Urban Development database [32] includes
data on Section 8 and low-income housing. These housing units
are a feature of built environments associated with known and
suspected health risks and disamenities.
The EPA's National Walkability[36] data is the source of
the walkability index. It combines data from 2010 Census
TIGER/Line shapefiles, 2010 Census Summary File 1, Census
Longitudinal Employer-Household Dynamics (LEHD) 2010,
InfoUSA 2011, NAVTEQ NAVSTREETS 2011, General Transit
Feed Specification (GTFS) data for 228 transit agencies, and the
Center for Transit Oriented Development (TOD) Database 2012
to produce a block group score, which was aggregated to the
county level.
The Landcover data derive from the EPA's National Land
Cover Database (NLCD)[34], It represents land cover across
the contiguous 48 states, circa 2011. Each 30-m2 pixel has been
classified using a standard land cover classification scheme, and
some of these categories have been aggregated further according
to procedures outlined in EPA's Report on the Environment[40],
Data originally were processed and compiled by the Multi-
Resolution Land Characteristics Consortium (MRLC)[41],
a United States federal interagency group, based on Landsat
satellite imagery. These data are combined with NOAA's C-CAP
Land[35] cover county data to represent land cover for all 3143
counties.
Summary of Changes to 2006-2010 data sources from
original 2000-2005 EQI
Air Domain - No changes to data sources
Water Domain - One data source was added for 2006-2010
(SDWIS), and some variables developed from the WATERS
database for 2000-2005 were not used in 2006-2010.
Land Domain - One data source was eliminated for 2006-2010
(National Geochemical
Survey). One data source was added for 2006-2010.
Mine Safety and Health Administration (MSHA) Mines Data
Set (2006-2010)
Sociodemographic Domain - No data sources were eliminated
for 2006-2010. Two data
sources were added to the 2006-2010 EQI.
USDAERS Creative class data
2008 Presidential Election results data
Built Domain - No data sources were eliminated for 2006-
2010. Two data sources were added to the 2006-2010 EQI.
EPA National Walkability data
EPA NLCD +C-CAP data
Variable Construction
Approach
We followed the same approach in developing variables for EQI
2006-2010 that we used for EQI 2000-2005. Most variables
throughout the different domains were identified previously and
developed as part of the EQI 2000-2005, then were updated for
the 2006-2010 period. For the newly added data sources, we
developed new variables. We assessed all variables as to whether
the new variables needed to be standardized, as a proportion
of geographical space (e.g., road proportions) or as a rate per
population (e.g., violent crimes per capita) for use in the EQI.
Additionally, some data were not available for all counties but
required spatial kriging to provide national coverage. Kriging is
a geospatial technique that uses known data points to interpolate
data at locations with unknown measurements [42],
The overall process for variable development for 2006-2010 was
as follows.
Update or identify and develop relevant variables within each
domain for each available year (2006-2010)
Assess collinearity among the variables within each domain
and eliminate redundant variables
Assess missing data and variability of each variable
Assess normality of variables and transform as necessary
Appendix II lists all the variables included in the EQI for each of
the five domains for 2006-2010 and includes notes about whether
the variables were used in the previous version of the EQI or
if they are newly created variables. Appendix III provides the
variables that were used in EQI 2000-2005 but were not used in
the EQI 2006-2010 update. The created variables are available
publicly at EPA's Environmental Dataset Gateway.
13
-------
Identification and Construction of Variables from Data
Sources
For each domain, all variables from EQI2000-2005 were
reviewed and assessed for continued inclusion in the EQI
2006-2010. Variables were created from selected data sources to
represent the constructs. Variables were developed in a variety
of manners, including kriging and standardization by area or
population. Each domain section below provides the details of
variable construction.
Assessing Variables
The data reduction method Principal Component Analysis (PCA)
is based on the variability between variables[43]; therefore,
collinearity of variables was assessed. This assessment was
done by developing correlation matrices for each domain.
Variables with any correlation coefficient >0.70 were examined;
representative variables were chosen for each pair or group of
highly correlated variables (Appendix IV).
Ideally, developed variables would have measured or estimated
values for each county of the United States. When this criterion
was not met, or when a majority (>50%) of values were zero,
the proportion of missing data and zero values were evaluated
for variable inclusion. If a particular variable had information
missing for many counties, the nature of the missing data was
evaluated. When it was determined that the missing data could
be interpreted as meaningful zeros (i.e., no measures were taken
because that condition did not occur in that county), the missing
values were set to zero. For instance, the counties with no
reported public housing were set to zero because public housing
is truly absent from some counties. When counties were missing
data because reporting areas were centralized, but the data could
not be assumed to be truly missing, the data were spatially
kriged when possible. For instance, crime was reported only for
specific counties, even though it likely occurred in counties other
than those in which it was reported as well. Therefore, crime
rates were averaged spatially over adjacent counties to create
an estimate for a county with no official reported crime. If the
missing data could not be determined to be legitimate zeros, the
data could not be reasonably kriged or averaged over geography,
and the number of counties with missing data was too high (more
than 50% of counties), the variable was not used in the EQI.
In some instances, there may have been more than one data
source that could represent a particular domain construct. In
that case, the data source deemed to have better data quality and
coverage was used.
Finally, normality of variables was evaluated. Using PCA,
the chosen data reduction technique, a key assumption is that
variables are distributed normally[43]. If data were nonnormal,
transformations were applied (typically log-transformation) to
increase normality. For those variables with zero values, half of
the nonzero minimum value was added to all observations before
log-transformation.
When data were updated on an annual or regular basis, variable
consistency (mean and standard deviation) was compared across
each year of the 5-year period (2006-2010).
Summary of Activities
Domain-Specific Variable Descriptions
Air Domain
The air domain consists of two data sources, (1) the AQS[10] and
(2) the NATA[11], representing criteria air pollutants and HAPs.
Criteria Air Pollutants
Daily concentration data from the EPA's AQS monitors (point
scale) were downloaded for ozone, carbon monoxide (CO), sulfur
dioxide (S02), nitrogen dioxide (N02), particulate matter under
10 |im in aerodynamic diameter (PM10), and particulate matter
under 2.5 |im in aerodynamic diameter (PM2.5). Annual averages
were calculated for each of the six pollutants at each monitor
with data. These averages then were used in a kriging procedure
to estimate annual concentration at each county's center point for
each year from 2006 to 2010.
For the EQI spanning 2006 to 2010., a single average
concentration was calculated from the annual average
concentrations for each county from the kriged estimates. When
indicated (i.e., lognormal distribution) half of the minimum
nonzero value was added, and variables were log transformed.
Hazardous Air Pollutants (HAPs)
County-level concentrations estimates from NATA were used
for all HAPs included in the EQI. HAPs were selected for
inclusion from the full NATA pollutant list. Using data from
2005, variables were evaluated for collinearity and variability.
Variables with any correlation coefficient >0.70 were examined,
and representative variables were chosen for each pair or group
of highly correlated variables (see Appendix IV). Correlations
were determined after assessing for missingness/zeros and
assessing normality. The variable that is correlated with the most
other variables was chosen. For example, if variable A was highly
correlated with variables B, C, D, and E, but each of those were
correlated with a lower number of variables, A was chosen as the
representative variable. The nonchosen variables (B, C, D, and E)
then were removed from consideration within other groupings.
If the correlation group was isolated (i.e., no variables in it were
associated with any other variables outside the isolated group),
then a representative variable was chosen without particular
criteria. By the end, all variables remaining had correlation less
than 0.7 with each other. All variables excluded were highly
correlated with (represented by) at least one variable that was
retained. Of the remaining variables, all missing values were
set to zero, with the assumption that lack of estimate for an area
indicated low concern for contamination with a particular HAP,
and the number of zero values was evaluated for each variable.
Pollutants with more than 50% zero values were dropped. This
process left 37 HAPs included in the EQI. When indicated (i.e.,
log-normal distribution), half of the minimum nonzero value was
added, and variables were log transformed.
14
-------
Table 3. 2005 NATA variables included in EQI 2006-2010
1,1,2,2-tefrachloroethane
1,1,2-trichloroethane
1,2-dibromo-3-chloropropane
1-3-dichloropropene
Acrylic acid
Benzidine
Benzyl chloride
Beryllium compounds
bis-2-ethylhexyl phthalate
Carbon tetrachloride
Carbonyl sulfide
Chlorine
Chlorobenzene
Chloroform
Chloroprene
Chromium compounds
Cobalt compounds
Cyanide compounds
Dibutylphthalate
Ethyl benzene
Ethyl chloride
Ethylene dibromide
Ethylene dichloride
Formaldehyde
Glycol ethers
Hydrazine
Hydrochloric acid
Isophorone
Manganese compounds
Methyl bromide
Methylene chloride
Phosphine
Polychlorinated biphenyls
Propylene dichloride
Quinoline
Trichloroethylene
Vinyl chloride
The air domain includes 43 variables representing criteria and
HAPs.
Water Domain
The water domain included six data sources: (1) the WATERS
program database[12], (2) Estimated Use of Water in the United
States[14], (3) the National Atmospheric Deposition Program
(NADP)[13], (4) the Drought Monitor Network[ 15], (5) the
National Contaminant Occurrence Database (NCOD)[16], (6)
the Safe Drinking Water Information System (SDWIS)[17]
Using these six data sources, variables were created to represent
seven constructs that describe the overall water environment.
The seven constructs were (1) overall water quality, (2) general
water contamination, (3) drinking water quality, (4) domestic
use, (5) atmospheric deposition, (6) drought, and (7) chemical
contamination.
Overall Water Quality
Impairment and water quality standards (WQS) data were
obtained for the most recent state reported data that were
collected under Sections 303(d) and 305(b) of the Clean Water
Act (CWA)[44], The CWA is administered at the state level,
and data are reported voluntarily from the states to the federal
level. The dates of the reported data ranged from 2004 to 2010,
as the federal reporting system maintains only the most recent
data reported by each state. Under Section 305(b) of the CWA,
states establish WQS for each hydrological feature based on the
expected use (or uses) of these waters. Under Section 303(d)
of the CWA, states assess whether waters are impaired (do not
meet the standards) for the uses established in the WQS. This
assessment is conducted biennially, and the states voluntarily
report these data to the federal level.
County-level impaired stream length was estimated for the
contiguous United States using impairment and WQS data (from
the WATERS database). With the designated uses listed for
each state, the WQS was classified into five broad categories of
water use: (1) agriculture, (2) drinking water, (3) recreation, (4)
wildlife, and (5) industry. Using geographic information systems
(GIS), county-level percentages of impairment were calculated.
WQS and impairment datasets were joined to the map layer of
hydrologic features in EPA's RAD [3 9], RAD is a replicate of the
National Hydrography Dataset Plus [3 8] augmented for reporting
water quality data. The defined broad water use categories were
joined to the WQS data, and a table summarizing hydrologic
features with multiple uses was created. WQS and impairment
tables were assigned to features in the RAD using GIS Network
and Event tools. These tools link tabular database information
with linear or polygon features. Stream lengths were clipped by
county boundaries to calculate percent impairment by county.
Only linear water features were included in each category.
Polygon features, such as lakes, were excluded because of the
lack of well-defined county and state boundaries across water
bodies. Next, county and state designations were linked with
linear features in RAD. Once all data were associated to linear
hydrologic features, lengths were calculated for water features
impaired for any use, drinking water use, or recreational use and
for all stream lengths within a county. The final variable was
cumulative measure of percent of water impaired for any use.
General Water Contamination
Water contamination can be caused by several sources.
Unfortunately, EPA only has consistent data on the point sources
of contamination in the form of the number of National Pollutant
Discharge Elimination System (NPDES)[45] permits. Therefore,
the number of permits in a county was used as a proxy for
general water contamination. Using permit information in the
WATERS database, 13 variables were calculated for the number
of discharge permits in a county. Permits that were current
during the period 2006-2010 were selected. The 10 variables that
were calculated based on individual permit types had too many
missing data; therefore, three composite variables were created
for inclusion in the EQI. A composite variable was developed
for the number of sewage permits per 1000 km of stream
length in a county. The number of animal feeding operations
and concentrated animal feeding operations NPDES permits,
combined sewer overflow NPDES permits, and NPDES permits
15
-------
for sludge in each county were summed and divided by the
total stream length in the county. Similarly, composite variables
were calculated for industrial permits (combining the total of
pretreatment NPDES permits, general facilities NPDES permits,
and individual facilities NPDES permits) and stormwater permits
(combining the total of general stormwater NPDES permits,
industrial stormwater NPDES permits) by county per 1000 km of
stream length. Preliminary analyses demonstrated low loadings
for the grouped variables; therefore, only one variable was
maintained: the total number of discharge permits per 1000km of
stream length in the county.
Drinking Water Quality
In the United States, drinking water quality is measured and
maintained by the public water system (PWS) treating and
distributing drinking water. Based on the Safe Drinking Water
Act, states are required to report basic information about
PWS, violation information for each PWS, and enforcement
information to the federal system. The SDWIS data is publicly
available data through the Fed Data Warehouse [17], The basic
information for the PWSs were merged with the violations
reports, so that the county and city served by the violations were
together in one report. In instances where there were multiple
counties served by a PWS, the counties were separated to account
for these violations in both counties served by the PWS. Variables
were created for each rule within the Safe Drinking Water Act,
such as the Lead and Copper Rule. A time period average for
each rule name violation by PWS was calculated as the frequency
divided by the number of years in the time period of interest, in
this case five (2006-2010). This time period average was then
multiplied by the population served for each PWS, and these
values were summed for the county to estimate the proportion
of the population in the county affected by the violation. Most
counties did not report violations for the majority of rules;
therefore, only one variable constructed provided sufficient
variability to be included, which was that calculated from
violations to the Total Coliform Rule.
Domestic Use
Data from the Estimated Use of Water in the United States
database[14] were used as a proxy for domestic water quality.
If water is being withdrawn for competing uses (agriculture,
industry, etc.), it will put stress on water supplies, which, in turn,
will affect water quality. This database includes county-level
estimates of water withdrawals for domestic, agricultural, and
industrial use. Initially, 15 variables of water withdrawals for
domestic, agricultural, and industrial use were developed. These
data are estimated every 5 years and were included in the EQI as
averaged data for 2006 and 2010. Two variables were included in
the EQI after evaluation for collinearity (four variables removed)
and missing data (nine variables removed). The two variables
were (1) the percent of population on self-supplied water
supplies and (2) the percent of those on public water supplies
that are on surface waters. For these variables, higher values are
not necessarily a marker for poor water quality. The data were
provided at the county level and normally distributed; therefore,
no additional transformation was required.
Atmospheric Deposition
The atmospheric deposition of chemicals can affect water quality.
The NADP dataset[13] provides measures for the concentration
of nine chemicals in precipitation: (1) calcium, (2) magnesium,
(3) potassium, (4) sodium, (5) ammonium, (6) nitrate, (7)
chloride, (8) sulfate, and (9) mercury. Annual summary data
from each monitoring site for each year 2006-2010 were kriged
spatially to achieve national coverage and county-level estimates.
The annual estimates for each pollutant then were averaged
over the 5-year study period. The data for all pollutants, except
sulfate, were skewed and, therefore, were natural log transformed
to achieve normal distributions. Magnesium, sodium, and
ammonium were removed as they were highly correlated with
potassium, chloride, and nitrate, respectively.
Drought
Drought affects the concentration of pathogens and chemicals in
water bodies and, therefore, can affect water quality. The Drought
Monitor dataset[15] provides raster data on six possible drought
status conditions for the entire United States on a weekly basis.
The data were aggregated spatially to the county level to estimate
the percentage of the county in each drought status condition. The
weekly data were averaged to achieve annual estimates for 2006-
2010 and, then, averaged to create a composite for the entire
period. From this data, the percentage of the county in extreme
or exceptional drought (intensity levels D3 and D4, respectively)
was used in the EQI. The remaining five drought status
conditions were removed because all of the drought statuses were
highly correlated.
Chemical Contamination
Chemical contamination of water supplies can directly affect
human health. The NCOD dataset[16] provides data on 69
contaminants provided by public water supplies throughout the
country for the period from 1998-2005. Data for all samples
in a county for each contaminant were averaged over the
entire period of the dataset, 1998-2005. More recent data were
not available. The data also were natural log transformed to
achieve normal distributions. Missing values were set to zero,
with the assumption that lack of measurement for an area
indicated low concern for contamination with that particular
contaminant. Nine contaminants, (1) asbestos, (2) beryllium,
(3) diquat, (4) endothall, (5) glyphosate, (6) dioxin, (7) radium,
(8) beta particles, and (9) uranium, did not include data for
enough counties (missing data) to be included in the EQI
construction. Twenty-one variables were deleted because of
high correlation with other contaminants: (1) lindane, (2)
thallium, (3) toxaphene, (4) oxamyl, (5) alachlor, (6) 2,4,5-TP
(Silvex), (7) hexachlorocyclopentadiene, (8) carbofuran, (9)
heptachlor, (10) heptachlor epoxide, (11) hexachlorobenzene,
(12) 1,2,4-trichlorobenzene, (13) 1,2-dichlorobenzene,
(14) vinyl chloride, (15) 1,1-dichloroethylene, (16)
trans-1,2-dichloroethylene, (17) 1,2-dichloroethane,
(18) carbon tetrachloride, (19) 1,2-dichloropropane, (20)
1,1,2-trichloroethane, (21) benzene.
Land Domain
The land domain consisted of five data sources, representing five
constructs: (1) agriculture, (2) pesticide use, (3) facilities, (4)
radon zone, and (5) mining activity.
16
-------
Agriculture
Information on nonpesticide chemicals used in farming, animal
units, harvested acreage, irrigated acreage, manure acreage,
and proportion of farms was taken from the 2007 Census of
Agriculture [19], Final acreage for each item then was divided
by total acreage for each county to return a percentage (e.g.,
percentage of irrigated acres out of total acres in a county). In
some cases, county-level acreage for items was suppressed. In
these, case estimates were imputed based on unaccounted for and
total state-level acreage. Known acreage was subtracted from
total state acreage, leaving an "unassigned" total acreage for
each state. This total number was divided by the total number of
farms in counties with suppressed acreage to return an average
acreage for each farm. This average acreage then was multiplied
by the number of farms in each county with suppressed acreage
to estimate acreage. Animal units were estimated by multiplying
the number of livestock (cows, hogs, and poultry) by the animals
per animal unit statistic [46] and, then, adding together all
livestock categories for each county. Eight variables representing
agriculture were included in the EQI.
Pesticide Use
Pesticide use for each county was estimated using county-
pesticide-use data from the 2009 National Pesticide Use
Dataset[18], Each pesticide was categorized into one of three
categories: (1) herbicide, (2) fungicide, or (3) insecticide. The
average weight (in kilograms) of each pesticide was calculated
for the years available (2006-2009) for each county, then
summed by pesticide type. If a county did not have information
for one of the pesticide categories, the national average was
used. Despite the choice of high spatial coverage, there are
recognized uncertainties in estimating the geographic distribution
of compounds applied to specific crops as described by Baker et
al. (2015) in prior literature [47], These three pesticide categories
were included in the EQI. Pesticide variables were evaluated for
normality and log transformed.
Facilities
Large facilities have the capacity to affect land quality. The
facilities included in the land domain are those represented on
the EPA Geospatial Data Download Service [20], Because many
counties had at least one, but no counties had all six of the
facility types present, a composite facilities data variable was
constructed by summing the count of any one of the six facilities
types (Brownfield sites (n=1273)[48]; Superfund sites (n=719)
[49]; Toxic Release Inventory sites (n=2671)[20]; pesticide-
producing-location sites (n=2099)[50]; large-quantity generator
sites (n=1963)[51]; and treatment, storage, and disposal sites
(n=874)[52]) across the counties. Facilities were included in the
count if they were identified during the 2006-2010 period. The
count of facilities was divided by the county population, which
produced a facilities rate. The facilities rate variable was assessed
for normality and log transformed.
Radon Zone
The potential for elevated indoor radon levels was represented
using the county score from the EPA Radon Zone map[21],
which was available for 3142 counties (one county, Broomfield,
Colorado, was missing). The EPA Radon Zone map identified
areas of the United States with the potential for elevated indoor
radon levels. Each United States county was assigned to one of
three zones based on radon-level elevation potential.
Mines
Mines, like large facilities, have the capacity to affect land
quality. The mines included in the land domain are those found
in the MSHA dataset[22], which includes those mines under
MSHA jurisdiction since 1970. Mines were included if they were
active at any point before 2010 and were not abandoned and
sealed after 2006. Those excluded most likely do not continue to
pose any environmental impact. Any mines already represented
in Superfund data were excluded. Mines were separated by the
five primary commodity types: (1) coal, (2) metal, (3) nonmetal,
(4) sand and gravel, and (5) stone, and a county could have more
than one type of mine. The counts of the mines were divided
by the county population, producing a mine rate. Of the 3143
counties, 2904 had at least one mine. For those counties that
had zero values for the different mine types, zeros were replaced
with the minimum value of the mine type/2 was added to the
standardized population variables. The mine variables were
assessed for normality and log transformed.
Sociodemographic Domain
This domain was constructed to explore the sociodemographic
features of counties in the United States. These features were
used to approximate the social stress associated with residing in
more deprived (low education, high unemployment, high violent
crime, high poverty, etc.) or more affluent (high employment
rates, low property crime, high proportion of college graduates,
etc.) counties. This domain includes variables from the 2010
United States Census) [24], the FBI Uniform Crime Reports
(UCR)[25], the 2008 Presidential election results[26], and the
United States Department of Agriculture Economic Research
Service Creative Class data[27]. Because the sociodemographic
domain is related to population density, by virtue of the data's
collection and reporting, variables were developed as population
rates (denominator: count of persons per county), rather than
area-based rates (denominator: square miles per county).
Nine variables were obtained from the 2010 United States
Census [24], The nine variables were (1) percent earning a
bachelor's degree or higher among persons aged 25 years or
older; (2) percent persons unemployed; (3) percent of families
living below the federal poverty line; (4) percent vacant housing
units; (5) median household value; (6) median household income;
(7) percent renter-occupied units; (8) count of occupants per
room; and (9) the Gini coefficient, a marker of income inequality.
Owing to the skewed nature of the household income and count
of occupants per room data, these variables were log transformed
for inclusion in the EQI. The sociodemographic domain contains
a mix of positive and negative features; therefore, when the
sociodemographic domain was constructed, positive variables
were reverse-coded to ensure that a higher amount of the
sociodemographic domain will represent adverse environmental
conditions.
The area-level crime environment was represented using the
FBI UCRs[25], The first step in constructing crime data was to
assign each jurisdiction or place to a county using county Federal
Information Processing Standards code[53]. In cases when a
jurisdiction covered more than one county, the reported crime
17
-------
was assigned to both counties. Although this double assignment
results in a slight inflation of crime reports for a state, there was
no way to determine which county should receive the crime
report. Further, if police or municipal jurisdictions crossed county
lines, it is likely residents of both counties were "exposed" to
the crime environment. Crime data attributed to more than one
county occurred in approximately 15 counties. Second, because
crime was reported for less than half the United States counties,
crime data were kriged spatially and temporally to estimate
values for counties with no reported crime. The decision was
made to krig these data because data reporting was voluntary,
and it seemed unlikely that no crime occurred in the nonreported
areas. Because zeros could not be reasonably assigned to the
missing counties, the data were interpolated spatially and
temporally instead. Based on experience with the 2000-2005
county-level EQI, and in acknowledgement that the correlation
between the property and violent crime rates was very high
(0.96), only log violent crime was included in the EQI.
The political climate of a county was represented by Leip's
election map [26], On this Web site, county-specific percents
voting Republican or Democratic are reported. These data were
downloaded for each county. The report voting Democratic in the
2008 presidential election are included in the EQI. One county in
Hawaii that had been an independent county unit, FIPS 15005,
was subsumed by Maui for the presidential election data, so the
same Democratic percentage was applied to county 15005 as to
Maui.
One creative class variable was included in the 2006-2010 EQI.
The creative class thesisthat towns need to attract engineers,
architects, artists, and people in other creative occupations to
compete in today's economymay be particularly relevant to
rural communities, which tend to lose much of their talent when
young adults leave. The ERS creative class codes[27] indicate
a county's share of population employed in occupations that
require "thinking creatively." The percent employed in creative
class occupations index was included in the EQI.
Built Domain
Seven data sources were included in the built domain,
representing (1) the subsidized housing environment, (2) traffic
safety, (3) public transportation usage and commuting times,
(4) road properties (road type and density), (5) the business
and service environments (e.g., food, recreation), (6) county
walkability, and (7) green space.
Housing Environment
The subsidized housing environment was represented by the
Housing and Urban Development data[32]. These data provide
a count of the low-rent and Section 8 housing in each housing
authority data area. The housing authority areas correspond to
cities, which were assigned county codes. Data were collected
in 2010, but, because low-rent and Section 8 housing does not
change substantially over time, these data were considered
representative of the 2006-2010 period. The variables were
summed to result in the count of any low-rent or Section 8
housing in each county. The rate of subsidized housing was
constructed by dividing the count of subsidized housing units per
county by the county population. The data were log transformed
prior to inclusion in the EQI.
Traffic Safety
Traffic fatalities, an important feature and consequence of
the built environment, were estimated using the Fatality
Analysis Reporting System (FARS) data[31], The FARS is a
national census providing the National Highway Traffic Safety
administration yearly reports of fatal injuries suffered in motor
vehicle crashes. Rates for the 2006-2010 counts of fatal crashes
per county were constructed by dividing the count of county-level
fatal crashes by the county-level population. Many counties had
no fatal crashes. To accommodate the large number of meaningful
zeros in the data, the log of this rate variable was used in the built
domain of the EQI.
Public Transportation Usage and Commuting Time
The percent of county residents who use public transportation
was estimated using the 2010 United States Census[24] variable
in the EQI. For many counties, the percent of the population that
reports using public transportation is near zero. Therefore, this
variable was log transformed prior to its use in the built domain
of the EQI. Also obtained from the United States Census was
the average number of minutes employed persons spent on the
commute home from work.
Road Properties
For the built-environment domain, characterizing the relative
proportions of each county that was served by highways,
secondary roads, and primary roads were of interest, as these
types of roads confer different risks (related to speed and
safety) and benefits (related to neighborhood walking or ease of
transit). Road type for the year 2008 was approximated using the
NAVTEQ road data[30] associated to TIGER county boundary
[29] data. Three proportion variables were constructed by
dividing the mileage of each road type (e.g., secondary roads)
by the total road mileage in each county. The proportions of all
roadways that were secondary roads were included.
Business and Service Environments
Businesses represent an important component of the built
environment and can contribute to the risk and amenity
landscape. Variables representing various built-environmental
features were constructed using the proprietary 2008 Dun and
Bradstreet data[28], which include commercial information on
businesses and data on more than 195 million records. Eight rate
variables were constructed by dividing the county-level count of
a business type by the county-level population count. The eight
variables included the (1) positive food environment, (2) negative
food environment, (3) vice environment (alcohol, pawn, and
gaming), (4) health care business environment, (5) recreation
environment, (6) education environment, (7) social-service
environment, and (8) civic-related environment. Note: Positive
food environments included those that sold healthier foods,
like grocery stores, sit-down restaurants, and organic shops,
whereas the negative food environment included businesses like
fast-food restaurants, convenience stores, and pretzel trucks.
Although related, these two food environments comprise different
businesses and are not 100% inversely correlated. Nonnormally
distributed variables were log transformed, and all eight were
included in the EQI.
18
-------
Walkability
Walkability is an important feature of the built environment, and
variability across walkability may help explain poor or good
health. The National Walkability Index (NWI)[36] was used to
determine walkability as a mode of travel for each county. The
scores, ranging from zero to 20 are calculated using a weighted
rank of four variables: (1) mix of employment types (such as
office, retail, and service) and occupied housing, (2) mix of
employment types in a block group (such as office, retail, and
service), (3) street intersection density (pedestrian-oriented
intersections), and (4) predicted commute mode split - proportion
of workers in the block group who carpool. A higher rank
indicates an increased likelihood of walking being used as the
mode of travel. The block group scores were added, and, then,
a mean of the block group scores based on county population
proportions was created. The county walkability scores ranged
from 1.00 to 16.23.
Green Space
Exposure to green space also has been associated with
improved health. The green space variable was created by EPA's
EnviroAtlas[33] using National Land Cover Database (NLCD)
[34] and Coastal Change Analysis Program[35] data. Three
possible constructions were considered: The NINDEX variable
was created by EnviroAtlas as a natural land cover variable
and includes barren land, forest, shrub/scrub, grassland, sedge,
lichens, moss, and wetlands. NINDEXopen is the NINDEX
variable with developed open space, such as parks and golf
courses, included. The Richardson index[54] is based on a green
space paper and includes the NINDEX and also developed
open space, low intensity, and medium intensity. For the sake of
dissemination outside academic communities and ease of data
availability/construction, the 2006-2010 EQI used the NINDEX_
open variable. The variables represented percentages of up to 24
possible land cover types. To create a green space variable, five
total land cover groups were combined, those classified as (1)
natural land cover (barren land, rock/sand/clay/tundra/perennial
ice), (2) forest, (3) shrubland/scrub land, (4) herbaceous, and (5)
wetlands) and those classified as developed open space, where
impervious surfaces make up less than 20% of total cover and
includes recreational areas, such as grassy lawns, parks, and
golf courses. This combined variable of natural land cover and
developed open space gave a percentage of the county that had
green space and ranged from 3.88% to 99.99%. The variable then
was assessed for normality.
Changes to 2006-2010 variable construction from
original 2000-2005 EQI
Air Domain
Variables eliminated from the 2006-2010 EQI
The following air variables were eliminated because of high
collinearity to one or more variables.
Variable Represented by
2-4-toluene diisocyanate Ethylbenzene
2-chloroacetophenone Benzyl chloride
2-nitropropane Chloroprene
4-nitrophenol
Ethylbenzene
Acetophenone
Ethylbenzene
Acrolein
Ethylbenzene
Acrylonitrile
Chloroprene
Biphenyl
Ethylbenzene
Bromoform
Benzyl chloride
Cadmium compounds
Chromium compounds
Carbon disulfide
Ethylbenzene
Cresol cresylic acid
Ethylbenzene
Cumene
Ethylbenzene
Diesel engine emissions
Ethylbenzene
Dimethyl formamide
Ethyl chloride
Dimethyl phthalate
Ethylbenzene
Dimethyl sulfate
Benzyl chloride
Epichlorohydrin
Chloroprene
Ethyl acrylate
Chloroprene
Ethylene glycol
Ethylbenzene
Ethylene oxide
Ethylene dichloride
Ethylidene dichloride
Vinyl chloride
Hexachlorobenzene
Polychlorinated biphenyls
Hexachlorobutadiene
Chloroprene
Hexachlorocyclopentadiene
Chloroprene
Hexane
Ethylbenzene
Lead compounds
Chromium compounds
Mercury compounds
Ethylbenzene
Methanol
Ethylbenzene
Methyl chloride
Carbon tetrachloride
Methyl isobutyl ketone
Ethylbenzene
Methyl methacrylate
Ethylbenzene
Methylhydrazine
Benzyl chloride
MTBE
Ethylbenzene
Nitrobenzene
Chloroprene
n-n-dimethylaniline
Chloroprene
o-toluidine
Chloroprene
PAH/POM
Ethylbenzene
Propylene oxide
Chloroprene
Selenium compounds
Ethylbenzene
Styrene
Ethylbenzene
Tetrachloroethylene
Ethylbenzene
Toluene
Ethylbenzene
Triethylamine
Ethylbenzene
Vinyl acetate
Ethylbenzene
Vinylidene chloride
Ethylbenzene
19
-------
Water Domain
New variables added to the 2006-2010 EQI
Total coliform health-based violations added
Variables removed in the recreational water construct
Number of days closed per event in county 2000-2005
numDays_Close_Activity_tot
Number of days per contamination advisory event in county
2000-2005
numDays_Cont_Activity_tot
Number of days per rain advisory event in county 2000-2005
numDays_Rain_Activity_tot
Variables removed in the chemical contamination construct from
the 2006-2010 EQI because of correlation with other variables
Beryllium - W Be ln (mg/L)
Lindane - W Lindane ln (mg/L)
Thallium - W_Tl_ln (mg/L) 1996
Toxaphene - W Toxaphene ln (|ig/L)
Oxamyl (Vydate) - W Oxamyl ln (|ig/L)
Alachlor - WAlachlorln (|ig/L)
2,4,5-TP (Silvex) - W silvex ln (|ig/L)
Hexachlorocyclopentadiene - W HCCPD ln (ng/L)
Carbofuran - WCarbofuranln (ng/L)
Heptachlor - WHeptachlorln (ng/L)
Heptachlor Epoxide - W Heptachlor epox ln (|ig/L)
Hexachlorobenzene - W HCB ln (ng/L)
1,2,4-Trichlorobenzene - W_124TCIB_ln (|ig/L)
1,2-Dichlorobenzene (o-Dichlorobenzene) - W ODCB ln
(lig/L)
Vinyl Chloride - W VCM ln (ng/L)
1,1-Dichloroethylene - W llDCE ln (ng/L)
trans-l,2-Dichloroethylene - W_tl2DCE_ln (ng/L)
1,2-Dichloroethane (Ethylene Dichloride) - W EDC ln (|ig/L)
Carbon Tetrachloride - W_CC14_ln (ng/L)
1,2-Dichloropropane - W PDC ln (ng/L)
1,1,2-Trichloroethane - W_112TCA_ln (ng/L)
Benzene - W Cllbenz ln (ng/L)
Land Domain
Variables eliminated from the 2006-2010 EQI
The following variables were eliminated because content was
represented in the NCOD and NADP.
Mean level of arsenic
Mean level of selenium
Mean level of mercury
Mean level of lead
Mean level of zinc
Mean level of copper
Mean level of aluminum
Mean level of sodium
Mean level of magnesium
Mean level of phosphourous
Mean level of titanium
Mean level of calcium
Mean level of iron
New variables added to the 2006-2010 EQI
Primarily coal mines per county population
Primarily metal mines per county population
Primarily nonmetal mines per county population
Primarily sand and gravel mines per county population
Primarily stone mines per county population
Sociodemographic Domain
Variables eliminated from the 2006-2010 EQI
Percent management occupation - eliminated because content
better covered in creative class index data
Housing built before 1939 - eliminated because of unclear
association with health
Percent with no English - eliminated because of unclear
association with health and increasing subjectivity
Variables substitutions for the 2006-2010 EQI
Percent bachelor's degree (>25 years old) substituted for
percent greater than high school
Percent family poverty substituted for percent persons in
poverty
Count of occupants per room replaced median number of
rooms
New variables added to the 2006-2010 EQI
Percent of persons working in creative occupations
Percent of county that voted Democratic in the 2008
presidential election
Built domain
Variables eliminated from the 2006-2010 EQI
Entertainment environment - eliminated because of unclear
association with health
Transportation environment - because the data contained in
this variable is better covered using other data sources
Variables substitutions for the 2006-2010 EQI
Percent secondary roads replaced percent primary roads
New variables added to the 2006-2010 EQI
Walkability score added
Proportion of county in green space added
20
-------
Data Reduction and Index Construction
Overall Approach
After variable development, all the variables were combined
into an index representing the overall environmental quality.
The specific tasks required for index construction were as
follows.
Included all the variables from one domain in a PCA to
empirically summarize that domain-specific environmental
context (retaining the first component as the domain index) for
each of the five domains
Assessed the positive/negative direction (valence) of the
variable loadings for each domain; if loadings were not in
the correct direction to ensure a higher value on the index
corresponded to worse environmental quality, corrected
valence when necessary
Combined each of the five domain-specific indices in another
PCA to empirically summarize the overall environmental
context into one index of environmental quality and retained
the initial component as the overall EQI
Repeated the three previous steps for each of the four RUCC
strata (e.g., RUCC stratum 1 air domain; RUCC stratum 2 air
domain, etc.), such that each RUCC had its own set of domain-
specific indices, as well as its own overall index
The EQI, domain-specific indices, and EQI stratified by rural-
urban data are available publicly at EPA's Environmental Dataset
Gateway. Also, an interactive map of the EQI is available at
EPA's GeoPlatform.
Principal Components Analysis (PCA)
PCA is a data reduction technique frequently used to create
sociodemographic scales or indices for inclusion in statistical
models[43, 55], PCA analyzes total variance, and the loading
represents the correlation between the variable and the
component. PCA assumes no underlying latent variable structure
but, rather, seeks to empirically summarize multiple possible
domains. Three major goals of PCA are to
1. summarize the patterns of correlations among observed or
measured variables,
2. provide an operational definitionin this case, a regression
equationfor underlying processes by using observed or
measured variables, and
3. reduce a large number of observed variables into a smaller
number of factors or a single component.
PCA was chosen for data reduction for several reasons.
Production of an empirical summary of the various constituent
components of the EQI was desired. Various data sources
measured on multiple scales needed to be combined. PCA
standardizes these measures prior to combining. Therefore,
the differing scales are less problematic. To assess variables
influences on the index, variables cannot simply be added
together. To do so would mean knowledge for most of the
variables would not be available to indicate if any one variable
would prove to be more "influential" for environmental quality
than another. PCA enables variable loadings to vary by their
relative importance to the total component. This feature enabled
exploration of variable loading differences for interpretation
purposes.
The PCA steps included
selecting the set of variables to be used,
preparing the correlation matrices,
extracting the set of components from the correlation matrix,
determining the number of components observed, and
interpreting the findings.
The sole modification to the PCA methodology in the county
2006-2010 EQI compared to that of the 2000-2005 EQI is
"valence correction." We also have created a 2000-2005 valence-
corrected version of the EQI.
"Valence correction" refers to reorientation of PCA output for
uniformity of interpretation of domain indices and uniformity
in orientation of domain indices input into the second PCA for
EQI construction. In this instance, we are defining valence as
the departure from neutrality along a continuum; generally, we
are interested how attributes depart from neutrality in opposite
directions. The PCA loadings are a function of the program's
starting point, or seed, which is not easily manipulable.
Therefore, the loading valence needed to be corrected prior to
the construction of the indices to ensure that higher values on a
given index, and on the overall EQI, signify worse environmental
quality [56, 57],
Domain and EQI indices are designed such that lower (more
negative) values represent "better" quality and higher (more
positive) values represent "worse" quality. Under this setup,
health beneficial variables should load negative in the PCA output
("+" or loading sign for a variable in the component variable
loadings vector represents positive or negative correlation
between that variable and the component, respectively). Given
that the first principal component was taken to represent domain
or environmental quality and that the orientation of these
indices was designated as going from better to worse quality
(negative to positive index value), it was necessary to reverse the
component variable loadings vector from a PCA output if a high
proportion of variables was deemed beneficial loaded "+", and
a high proportion of variables was deemed detrimental loaded
"-"[55], Determination of variables as beneficial or detrimental
to human health across domains was done a priori based on
literature evidence and content matter judgment. Reorientation
of PCA-derived indices through multiplication of the component
variables loading vector by -1 preserves (1) the direction of the
relationship among the variables for a given PCA (i.e., variables
that loaded with same signs will retain same signs, and variables
that loaded opposite to each other will retain opposite signs
after reversal, and, therefore, the pattern of correlations among
the variables will remain intact); and (2) the magnitude of
correlation among variables (reversal of loading signs does not
impact the magnitude of the loading) [5 8], The sum of squares
of variable loadings in a PCA output equals 1, and, therefore,
each square of a variable loading can be viewed as a measure of
the contribution of that variable toward the principal component
(domain indices and EQI in this case), enabling estimation of the
"correctness" of the orientation of the index. We used the square
21
-------
of variable loadings in a given PCA output in combination with
aforementioned a priori designations of benefit or harm to guide
choice of index reorientations.
PCA analyzes the total variance. Therefore, in the PCA
correlation matrix, "1" is In the positive diagonal[55]. To
construct the EQI. variables from each domain were entered into
domain-specific PC As. PCA produced variable loadings, which
were roughly equivalent to the "weight" or contribution that each
variable made toward explaining the total variance. The weights,
however, need not sum to 1.0 because the loadings were for the
total variance, rather than just the shared variance. The loading
associated with each variable then was multiplied by its mean
value for the given geography (county, for the EQI), and these
weighted mean values were summed.
Rural-Urban Continuum
Both the domain-specific indices and the overall EQI were
created for each county in the United States. Recognizing
that environments differ dramatically across the rural-urban
continuum[59], the decision was made that the EQI would be
most useful if it accommodated rural-urban environmental
differences. The EQI was stratified by RUCCs. The RUCC is
a nine-item categorization code of proximity to or influence of
major metropolitan areas[60]. The nine-item categories were
condensed into four, where RUCC1 represents metropolitan-
urbanized = codes 1+2+3, RUCC2 nonmetropolitan-urbanized =
4+5, RUCC3 less urbanized = 6+7, and RUCC4 thinly populated
(rural) = 8+9 (see Figure 3)[61-64], For the 2006-2010 EQI, the
2013 RUCC was used. RUCC-stratified EQIs, and an overall
EQI was constructed. Loadings on the stratified and nonstratified
sets of indices were assessed to determine loading heterogeneity
across counties. Because these loadings differed meaningfully
by RUCC level, RUCC-stratified EQIs were constructed
for each county.
Although it was possible to form as many independent linear
combinations as there were variables in PCA, only the first
principal component was retained. The first principal component
was the unique linear combination that accounted for the largest
possible proportion of the total variability in the component
measures. Therefore, the first component from each of these
domain-specific indices was retained (e.g., air index, water
index). Domain-specific indices were then entered into another
PCA, where the first component was retained as the EQI (Figure
2). This process was undertaken separately for each of the four
RUCC strata.
Within each RUCC strata, domain-specific variable loadings
were evaluated based on the value of variable loading and
the variable's hypothesized relevance to health. For instance,
although arsenic may occur in low frequency in a lot of counties
and, therefore, may have a relatively small component loading,
it is an important health hazard when present. Based on variable
loading magnitude alone, dropping arsenic from an EQI may be
a reasonable conclusion. However, it was retained for the EQI
based on its relevance to human health.
The first principal component, the domain-specific EQI (e.g.,
air domain EQI), then was standardized to have a mean of 0 and
standard deviation (SD) of 1 by dividing the index by the square
of its eigenvalue. Each domain-specific index was then included
in a second PCA procedure (Figure 2) to result in the overall EQI
for each stratum of RUCC.
For orientation to the results, low index scores (EQI and domain-
specific) indicate higher enviromnental quality, and higher index
scores (EQI and domain-specific) indicate lower enviromnental
quality.
Metropolitan urbanized
| Non-metro urbanized
| Less urbanized
| Thinly populated
Figure 3, Rural-urban continuum code (RUCC) stratification for all counties in the United States.
22
-------
Results
Description of Variables Comprising Environmental Quality
Index Domains
Air Domain
Criteria air pollutants were distributed relatively evenly
across the rural-urban gradient (Table 4). Some hazardous
air pollutants varied in emissions across rural-urban strata;
however, there was no discernable pattern for most. For example,
1,1,2-trichloroethane's highest levels were observed in the less
urbanized stratum, whereas levels were similar across other
strata, and emissions for manganese compounds were highest in
the most metropolitan areas then steadily decreased across more
rural strata.
Table 4. Air domain variable means, standard deviations (SDs), and ranges - Overall and rural-urban continuum codes (RUCCs) stratified
Metropolitan-
Urbanized ] Nonmetropolitan-
(RUCC1 = Urbanized (RUCC2 Less Urbanized Thinly Populated Total (3143)
1,167) Mean = 306) Mean (SD) (RUCC3 = 1,026) (RUCC4 = 644) Mean Mean (SD)
Variable Units (SD) [Range] [Range] Mean (SD) [Range] (SD) [Range [Range]
Construct: Criteria Air Pollutants
2.0E+01 (4.7E+00) 1.95E+01 (5.07E+00) 1.95E01 (4.37E+00) 1.89E+01 (4.88E+00) 2.0E+01 (4.7E+00)
PM10 (jg/m3 [4.1E-01,5.4E+01] [6.00E+00, 6.60e+01] [5.39E+00,5.25E+01] [4.01E-01,3.42E+01] [4.0E-01,6.6E+01]
1.1E+01 (2.1E+00) 1.02E+01 (2.19E+00) 9.99E+00 (2.20E+00) 9.05E+00 (2.39E+00) 1.0E+01 (2.3E+00)
PM2.5 |jg/m3 [4.1E+00,2.4E+01] [4.28E+00,1.48E+01] [3.35E+00,1.80E+01] [4.28E+00,1.79E+01] [3.3E+00,2.4E+01]
4.5E-02 (4.4E-03) 4.46E-02 (4.99E-03) 4.47E+02 (3.99E+03) 4.46E-02 (4.47E-03) 4.5E-02 (4.4E-03)
Ozone ppm [2.2E-02,5.9E-02] [2.22E-02,5.76E-02] [2.99E-02,5.72E-02] [2.90E-02,5.65E-02] [2.2E-02,5.9E-02]
9.2E+00 (4.6E+00) 7.93E+00 (3.93E+00) 3.85E-01 (8.36E-02) 6.65E+00 (4.37E+00) 8.0E+00 (4.4E+00)
Nitrogen oxide ppb [5.9E-01,3.1E+01] [5.92E-01,2.81E+01] [2.41E-01,8.89E-01] [5.91E-01,2.84E+01] [2.6E-01,3.1E+01]
2.2E+00 (1.5E+00) 1.97E+00 (2.22E+00) 7.53E+00 (4.00E+00) 1.47E+00 (1.39E+00) 1.9E+00 (1.5E+00)
Sulfur dioxide ppb [7.3E-03,9.7E+00] [1.10E-02,3.09E+01] [2.65E-01,2.84E-01] [2.21E-02,9.23E+00] [7.3E-03,3.1E+01]
3.9E-01 (8.2E-02) 3.87E-01 (7.49E-02) 4.32E-03 (4.91 E-04) 3.93E-01 (9.57E-02) 3.9E-01 (8.5E-02)
Carbon monoxide ppm [2.5E-01,8.7E-01] [2.49E-01,7.38E-01] [3.90E-03, 8.19E-03] [2.61E-01,8.90E-01] [2.4E-01,8.9E-01]
5.5E-04 (3.1 E-04) 5.47E-04 (3.47E-04) 5.50E-04 (3.14E-04) 4.77E-04 (2.75E-04) [ 5.4E-04 (3.1E-04)
Ethylene dibromide Tons emitted [5.5E-05,2.0E03] [1.65E-04,1.64E-03] [1.65E-04,1.79E-03] 5.50E-05,1.68E-03] [5.5E-05,2.0E-03]
1.9E+00 (6.0E-01) 1.75E+00 (5.57E-01) 1.79E+00 (5.80E-01) 1.61E+00 (6.05E-01) 1.8E+00 (6.0E-01)
Formaldehyde Tonsemitted [2.1E-01,5.6E+00] [6.83E-01,3.20E+00] [6.25E-01,3.86E+00] [2.08E-01,3.36E+00] [2.1E-01,5.6E+00]
4.4E-03 (7.5E-04) 4.46E-03 (9.07E-04) 1.39E-04 (2.79E-03) 4.20E-03 (6.61 E-04) 4.4E-03 (6.7E-04)
1,1,2,2-Tetrachloroethane Tonsemitted [1.3E-03,1.4E-02] [3.90E-03,1.33E-02] [1.76E-13,8.10E-02] [1.30E-03,1.60E-02] [1.3E-03, 1.6E-02]
4.0E-04 (6.6E-03) 2.00E-05 (1.24E-04) 5.25E-06 (9.53E-06) 9.61E-05 (1.58E-03) 2.1 E-04 (4.4E-03)
1,1,2-Trichloroethane Tonsemitted [1.8E-13,2.1E-01] [1.76E-13,1.73E-03] [1.95E-06,1.87E-04] [1.76E-03,3.59E-02] [1.8E-13,2.1E-01]
5.2E-06 (7.3E-06) 5.98E-06 (2.29E-05) 8.41 E-03 (2.26E-02) 4.34E-06 (6.27E-06) 5.1E-06 (1.0E-05)
1,2-Dibromo-3-chloropropane Tonsemitted [6.5E-07,9.1E-05] [1.95E-06,3.52E-04] [5.00E-16,3.75E-01] [6.50E-07,6.60E-05] [6.5E-07,3.5E-04]
1.1E-02 (3.4E-02) 1.06E-02 (2.13E-02) 6.41 E-05 (5.31 E-04) 5.00E-03 (1.38E-02) 9.1 E-03 (2.6E-02)
1,2-Dichloropropane Tonsemitted [5.0E-16,4.9E-01] [5.00E-16, 1.40E-1] [3.00E-015,1.01E-02] [5.00E-016,1.18E-01] [5.0E-16,4.9E-01]
1.4E-04 (2.4E-03) 2.06E-04 (2.45E-03) 3.43E-07 (7.89E-07) 9.76E-05 (1.39E-03) 1.1E-04 (1.8E-03)
Acrylic acid Tonsemitted [3.0E-15,7.2E-02] [3.00E-15,4.23E-02] [1.46E-08,7.29E-06] [3.00E-15,3.36E-02] [3.0E-15,7.2E-02]
3.3E-07 (1.2E-06) 3.22E-07 (1.98E-06) 1.26E-05 (2.92E-05) 3.14E-07 (1.60E-06) 3.3E-07 (1.3E-06)
Benzidine Tonsemitted [4.9E-09,3.6E-05] [1.48E-08,3.39E-05] [4.69E-12,3.90E-04] [4.88E-09,3.72E-05] [4.9E-09,3.7E-05]
1.4E-05 (3.9E-05) 1.40E-05 (4.08E-05) 1.26E-05 (2.92E-05) 1.10E-05 (4.97E-05) 1.3E-05 (3.9E-05)
Benzyl chloride Tonsemitted [4.7E-12,8.5E-04] [4.69E-12,4.20E-04] [4.69E-12,3.90E-04] [4.69E-12,1.16E-03] [4.7E-12, 1.2E-03]
4.4E-05 (4.4E-05) 4.55E-05 (6.00E-05) 4.66E-05 (8.23E-05) 3.57E-05 (2.93E-05) 4.3E-05 (5.9E-05)
Beryllium compounds Tonsemitted [7.5E-06,7.7E-04] [2.25E-05,6.93E-04] [2.25E-05,1.56E-03] [7.50E-06,6.26E-04] [7.5E-06,1.6E-03)
8.4E-03 (1.9E-03) 8.22E-03 (5.39E-04) 8.31 E-03 (1.77E-03) 8.08E-03 (6.40E-04) 8.3E-03 (1.6E-03)
bis-2-Ethylhexyl phthalate Tonsemitted [2.6E-03,6.3E-02] [7.80E-03,1.30E-02] [7.80E-03,4.36E-02] [2.60E-03,1.22E-02] [2.6E-03,6.3E-02]
9.1E-01 (1.8E-02) 9.11E-01 (3.75E-04) 9.11E-01 (9.67E-04) 9.06E-01 (5.36E-02) 9.1E-01 (2.7E-02)
Carbon tetrachloride Tonsemitted [3.0E-01,9.2E-01] [9.11E-01,9.15E-01] [9.03E-01,9.28E-01] [3.01E-01,9.27E-01] [3.0E-01,9.3E-01]
1.8E-03 (1.1E-02) 5.14E-03 (7.25E-02) 9.25E-04 (4.94E-03) 2.13E-03 (2.26E-02) 1.9E-03 (2.6E-02)
Carbonyl sulfide Tonsemitted [5.0E-16,1.6E-01] [5.00E-16, 1.27E+00] [5.00E-16,7.78E-02] [5.00E-16,4.39E-01] [5.0E-16, 1.35E+00]
2.4E-03 (1.9E-02) 3.25E-03 (2.48E-02) 1.57E-03 (9.72E-03) 1.34E-03 (8.28E-03) 2.0E-03 (1.6E-02)
Chlorine Tonsemitted [3.4E-13,5.6E-01] [3.41E-13,3.58E-01] [3.41E-13,1.76E-01] [3.41E-13,1.13E-01] [3.4E-13,5.6E-01]
4.2E-03 (1.5E-02) 3.40E-03 (1.17E-02) 2.73E-03 (9.33E-03) 1.60E-03 (5.08E-03) 3.1E-03 (1.1E-02)
Chlorobenzene Tonsemitted [3.4E-11,2.3E-011 [2.77E-07,1.63E-01] [1.01E-10,1.74E-01] [3.36E-11,5.42E-021 [3.4E-11,2.3E-01]
23
-------
Table 4. continued
Variable
Units
Metropolitan-
Urbanized
(RUCC1=
1,167) Mean
(SD) [Range]
] Nonmetropolitan-
Urbanized (RUCC2
= 306) Mean (SD)
[Range]
Less Urbanized
(RUCC3 = 1,026)
Mean (SD) [Range]
Thinly Populated
(RUCC4 = 644) Mean
(SD) [Range
Total (3143)
Mean (SD)
[Range]
Chloroform
Tons emitted
1.0E-01 (2.6E-02)
[3.0E-02, 6.6E-01]
9.77E-02 (1.61E-02)
[8.85E-02, 2.02E-01]
9.58E-02 (1.41E-02)
[8.85E-02, 2.26E-01]
9.36E-02 (1.31 E-02)
[2.95E-02, 2.11E-01]
9.7E-02 (2.0E-02)
[3.0E-02, 6.6E-01]
Chloroprene
Tons emitted
1.9E-04 (3.1E-03)
[1.6E-013, 8.8E-02]
1.06E-03 (1.81E-02)
[1.57E-13, 3.17E-01]
2.05E-04 (5.31E-03)
[1.57E-13,1.69E-01]
2.68E-05 (3.84E-04)
[1.57E-13, 7.24E-03]
2.4E-04 (6.7E-03)
[1.6E-13, 3.2E-01]
Chromium compounds
Tons emitted
4.1E-04 (7.0E-04)
[2.1E-05, 6.6E-03]
3.44E-04 (6.25E-04)
[6.15E-05, 5.63E-03]
3.28E-04 (7.70E-04)
[6.15E-05,1.04E-02]
2.18E-04 (4.00E-04)
[2.05E-05, 6.24E-03]
3.4E-04 (6.5E-04)
[2.1 E-05, 1.0E-02]
Cobalt compounds
Tons emitted
3.9E-05 (3.5E-04)
[2.2E-14, 8.5E-03]
2.66E-05 (1.12E-04)
[2.20E-14,1.66E-03]
2.91 E-05 (2.56E-04)
[2.20E-014, 6.95E-03]
3.80E-05 (2.92E-04)
[2.20E-14, 4.67E-03]
3.5E-05 (2.9E-04)
[2.2E-14, 8.5E-03]
Cyanide compounds
Tons emitted
2.5E-02 (6.1E-02)
[8.1E-14, 1.4E+00]
2.50E-02 (5.74E-02)
[8.10E-14, 8.76E-01]
1.76E-02 (2.15E-02)
[8.10E-014, 2.54E-01]
1.49E-2 (3.50E-02) [8.10E-
14, 8.00E-01]
2.1 E-02 (4.6E-02)
[8.1E-14,1.4E+00]
Dibutylphthalate
Tons emitted
3.5E-03 (5.3E-02)
[1.3E-09, 1.7E+00]
5.63E-03 (2.92E-02)
[3.81 E-08, 4.02E-01]
2.21E-03 (1 38E-02)
[7.18E-09, 2.19E-01]
1.76E-03 (2.94E-02)
[1.30E-09, 7.40E-01]
2.9E-03 (3.7E-02)
[1.3E-09,1.7E+00]
Ethyl chloride
Tons emitted
1.8E-03 (1.5E-02)
[7.6E-09, 5.1E-01]
1.18E-03 (1.67E-03)
[4.97E-08,1.31E-02]
1.42E-03 (9.95E-03)
[7.59E-09, 2.34E-01]
8.36E-04 (1.88E-03)
[7.59E-09, 2.93E-02]
1.4E-03 (1.1 E-02)
[7.6E-09, 5.5E-01]
Ethyl benzene
Tons emitted
7.7E-02 (1.2E-01)
[3.5E-05, 1.9E+00]
6.56E-02 (8.41E-02)
[1.78E-04, 5.41E-01]
5.88E-02 (8.87E-02)
[2.49E-04, 8.86E-01]
4.86E-02 (8.28E-02)
[3.46E-05, 8.46E-01]
6.4E-02 (1.0E-01)
[3.5E-05,1.9E+00]
Ethyl dichloride
Tons emitted
4.2E-03 (2.5E-03)
[9.0E-04, 3.9E-02]
4.17E-03 (3.10E-03)
[2.70E-03, 3.04E-02]
4.30E-03 (4.07E-03)
[2.70E-03, 7.73E-02]
3.89E-03 (4.38E-03)
[9.00E-04, 9.84E-02]
4.2E-03 (3.6E-03)
[9.0E-04, 9.8E-02]
Glycol ethers
Tons emitted
3.4E-03 (1.4E-02)
[1.8E-11, 2.5E-01]
2.68E-03 (8.45E-03)
[1.83E-11, 7.92E-02]
3.59E-03 (1 55E-02)
[1.83E-11, 2.66E-01]
2.63E-03 (1 35E-02)
[1.83E-11, 2.43E-01]
3.2E-03 (1.4E-02)
[1.8E-11, 2.7E-01]
Hydrazine
Tons emitted
4.2E-06 (1.4E-05)
[6.5E-08,1.4E-04]
4.60E-06 (1.46E-05)
[1.95E-07,1.21E-04]
3.27E-06 (1.25E-05)
[1.95E-07,1.83E-04]
3.34E-06 (1.67E-05)
[6.50E-08, 2.80E-04]
3.8E-06 (1.4E-05)
[6.5E-08, 2.8E-04]
Hydrochloric acid
Tons emitted
4.7E-01 (1.9E+00)
[3.7E-06, 2.5E+01]
2.08E-01 (1.04E+00)
[7.72E-05, 1.16E+01]
2.80E-01 (1.30E+00)
[1.11 E-05, 2.52E+01]
1.96E-01 (1.09E+00)
[3.69E-06, 2.15E+01]
3.3E-01 (1.5E+00)
[3.7E-06, 2.5E+01]
Isophorone
Tons emitted
1.1E-04 (9.4E-04)
[5.4E-14, 3.1E-02]
1.31E-04 (8.65E-04)
[5.40E-14,1.46E-02]
9.79E-05 (6.31 E-04)
[5.40E-14,1.71E-02]
4.55E-05 (1.63E-04)
[5.40E-14, 2.45E-03]
9.4E-05 (7.3E-04)
[5.4E-14, 3.1 E-02]
Manganese compounds
Tons emitted
2.4E-03 (1.8E-02)
[2.9E-04, 5.6E-01]
2.21E-03 (1.19E-02)
[8.70E-04, 2.03E-01]
1 58E-03 (3.79E-03)
[8.70E-04, 9.02E-02]
1.49E-03 (3.39E-03)
[2.90E-04, 6.50E-02]
1.9E-03 (1.2E-02)
[2.9E-04, 5.6E-01]
Methyl bromide
Tons emitted
6.8E-02 (5.2E-02)
[1.8E-02, 7.5E-01]
6.38E-02 (3.00E-02)
[5.25E-02, 2.90E-01]
6.19E-02 (3.21 E-02)
[5.25E-02, 5.94E-01]
5.77E-02 (1.66E-02)
[1.75E-02, 2.22E-01]
6.3E-02 (3.8E-02)
[1.8E-02, 7.5E-01]
Methyl chloride
Tons emitted
2.4E-01 (1.9E-01)
[5.5E-02, 4.7E+00]
2.31E-01 (1.29E-01)
[1.65E-01, 1.64E+00]
2.13E-01 (8.85E-02)
[1.65E-01,1.04E+00]
1.96E-01 (6.98E-02)
[5.50E-02,1.01E+00]
2.2E-01 (1.4E-01)
[5.5E-02, 4.7E+00]
Phosphine
Tons emitted
3.8E-05 (7.5E-05)
[2.6E-13, 8.3E-04]
3.72E-05 (6.85E-05)
[2.64E-13, 4.70E-04]
4.20E-05 (8.84E-05)
[2.64E-13,1.64E-03]
4.33E-05 (1.23E-04)
[2.64E-13, 2.59E-03]
4.0E-05 (9.1 E-05)
[2.6E-13, 2.6E-03]
Polychlorinated biphenyls
Tons emitted
3.8E-05 (1.1E-04)
[2.1E-13, 3.7E-03]
3.66E-05 (3.78E-05)
[2/06E-013, 2.99E-04]
3.14E-05 (3.47E-05)
[2.06E-013, 4.21E-04]
2.87E-05 (3.70E-05)
[2.06E-13, 4.88E-04]
3.4E-05 (7.4E-05)
[2.1E-13, 3.7E-03]
Propylene dichloride
Tons emitted
1.6E-03 (2.2E-03)
[2.3E-04, 4.5E-02]
1.21E-03 (1.06E-03)
[6.90E-04, 7.98E-03]
1.03E-03 (8.81 E-04)
[6.90E-04, 8.60E-03]
9.74E-04 (8.25E-04)
[2.30E-04, 7.00E-03]
1.3E-03 (1.6E-03)
[2.3E-04, 4.5E-02]
Quinoline
Tons emitted
1.4E-04 (2.7E-04)
[4.4E-07,1.7E-03]
1.51E-03 (3.27E-04)
[1.32E-06, 2.06E-03]
1.05E-04 (2.59E-04)
[1.32E-06,1.89E-03]
5.10E-05 (1 49E-04)
[4.40E-07,1.25E-03]
1.1 E-04 (2.5E-04)
[4.4E-07, 2.1E-03]
Trichloroethylene
Tons emitted
5.2E-02 (4.9E-02)
[2.5E-03, 7.6E-01]
4.69E-02 (4.06E-02)
[7.50E-03, 2.21E-01]
4.45E-02 (4.13E-02)
[7.50E-03, 2.84E-01]
3.48E-02 (4.08E-02)
[2.50E-03, 4.36E-01]
4.5E-02 (3.1E-03)
[2.8E-10, 7.0E-2]
Vinyl chloride
Tons emitted
7.8E-04 (3.8E-03)
[2.8E-10, 7.0E-02]
5.35E-04 (1.87E-03)
[2.84E-10, 2.35E-02]
6.01 E-04 (2.89E-03)
[2.84E-10, 5.59E-02]
4.55E-04 (2.64E-03)
[2.84E-10, 4.77E-02]
6.3E-04 (1.5E+00)
[7.3E-03, 3.1E+01]
24
-------
Water Domain
Variables included in the water domain (Table 5) suggest that
urban counties were more likely to have impaired stream length
(20%) compared with rural counties (9%). Additionally, urban
counties had higher mercury deposition, chloride precipitation,
sulfate precipitation, and the percentage of the county in drought
status. Chemical contamination varied by urban-rural status
depending on the chemical.
Table 5. Water domain variable means, standard deviations (SDs), and ranges - Overall and rural-urban continuum codes (RUCCs)
stratified
Variable
Construct: Domestic Use
Units
Metropolitan-Urbanized
(RUCC1 = 1,167)
Mean (SD) [Range]
Nonmetropolitan-
Urbanized (RUCC2
= 306)
Mean (SD) [Range]
Less Urbanized
(RUCC3 = 1,026)
Mean (SD)
[Range]
Thinly
Populated
(RUCC4 = 644)
Mean (SD)
[Range]
Total (3143)
Mean (SD)
[Range]
2.30E+01 3.62E+01
(3.83E+01) (4.23E+01)
4.47E+01 (4.29E+01) [0.00E+00, 4.35E+01 (4.24E+01) 3.26E+01 (4.13E+01) [0.00E+00, [0.00E+00,
Percent pop. on self-supply % 1.00E+02] [0.00E+00,1.00E+02] [0.00E+00, 1.00E+02] 1.00E+02] 1.00E+02]
3.38E+01 2.77E+01
(2.46E+01) (2.18E+01)
Percent pop. on self supply that 2.33E+01 (2.10E+01) [-2.62E-04, 2.40E+01 (1.72E+01) 2.99E+01 (2.10E+01) [-6.78E-02, [-6.78E-02,
is surface water % 1.00E+02] [0.00E+00, 8.20E+01] [-4.17E-02,9.21E+01] 1.00E+02] 1.00E+02]
Construct: Overall Water Quality
1.50E+01
9.02E+00 (2.05E+01)
Percent of stream length 1.97E+01 (2.35E+01) [1.00E-03, 1.72E+01 (2.30E+01) 1.28E+01 (1.84E+01) (1.33E+01) [1.00E- [1.00E-03,
impaired % 1.56E+02] [1.00E-03, 1.00E+02] [1.00E-03,1.08E+02] 03,1.00E+02] 1.56E+02]
Construct: General Water Contamination
4.74E+01
1.20E+01 (1.25E+02)
NPDES permits per 1000 km 9.08E+01 (1.91E+02) [1.00E-03, 3.44E+01 (3.90E+01) 2.42E+01 (4.34E+01) (2.40E+01) [1.00E- [1.00E-03,
of stream proportion 2.39E+03] [1.00E-03,2.97E+02] [1.00E-03,7.05E+02] 03,3.55E+02] 2.39E+03]
Construct: Atmospheric Deposition
2.23E-01 (1.09E- 1.90E-01 (1.1 IE-
Calcium precipitation weighted 1.63E-01 (9.69E-02) [1.22E-02, 1.83E-01 (1.10E-01) 2.03E-01 (1.20E-01) 01) [3.66E-02, 01) [1.22E-02,
mean mg/L 5.94E-01] [1.22E-02,7.48E-01] [3.80E-02,1.06E+00] 8.63E-01] 1.06E+00]
2.83E-01 (7.16E- 2.66E-01 (5.31E-
Potassium precipitation weighted 2.57E-01 (3.63E-02) [1.22E-01, 2.57E-01 (3.98E-02) 2.67E-01 (5.60E-02) 02) [1.58E-01, 02) [1.22E-01,
mean mg/L 4.91E-01] [1.22E-01,4.44E-01] [1.68E-01,1.01E+00] 1.11E+00] 1.11E+00]
7.55E-01 (2.07E- 7.42E-01 (2.10E-
7.34E-01 (2.11E-01) [0.00E+00, 7.38E-01 (2.40E-01) 7.44E-01 (2.03E-01) 01) [5.47E-03, 01) [0.00E+00,
Nitrate precipitation mg/L 1.13E+00] [0.00E+00,1.14E+00] [1.93E-02,1.14E+00] 1.14E+00] 1.14E+00]
1.88E-01 (1.77E- 2.44E-01 (2.13E
Chloride precipitation weighted 2.98E-01 (2.44E-01) [3.47E-02, 2.37E-01 (2.19E-01) 2.22E-01 (1.79E-01) 01) [7.19E-02, 01) [3.47E-02,
mean mg/L 1.91E+00] [3.47E-02,1.56E+00] [6.94E-02,2.15E+00] 1.58E+00] 2.15E+00]
1.03E+00
9.26E-01 (2.76E- (3.28E-01)
Sulfate precipitation weighted 1.10E+00 (3.39E-01) [1.00E-01, 1.05E+00 (3.78E-01) 1.02E+00 (3.10E-01) 01) [2.03E-01, [1.00E-01,
mean mg/L 1.89E+00] [1.00E-01,1.96E+00] [2.00E-01,2.09E+00] 1.92E+00] 2.09E+00]
9.15E+00
8.43E+00 (2.71 E+00)
9.44E+00 (2.59E+00) [2.81 E-02, 9.02E+00 (2.67E+00) 9.29E+00 (2.66E+00) (2.88E+00) [1.60E- [2.62E-02,
Total mercury deposition ng/m2 1.84E+01] [2.62E-02,1.76E+01] [3.62E-01,1.55E+01] 01,1.46E+01] 1.84E+01]
Construct: Drough
3.43E+00 3.84E+00
(5.92E+00) (6.75E+00)
Percent of county drought 4.16E+00 (7.38E+00) [0.00E+00, 3.70E+00 (6.67E+00) 3.76E+00 (6.51 E+00) [0.00E+00, [0.00E+00,
extreme % 4.52E+01] [0.00E+00,3.87E+01] [0.00E+00,4.82E+01] 4.43E+01] 4.82E+01]
Construct: Chemical Contamination
2.67E-03 (3.24E- 3.46E-03 (4.66E
3.59E-03 (5.10E-03) [1.00E-03, 3.61E-03 (3.53E-03) 3.75E-03 (5.13E-03) 03) [1.00E-03, 03) [1.00E-03,
Arsenic mg/L 1.34E-01] [1.00E-03,3.90E-021 [1.00E-03,7.20E-021 3.10E-021 1.34E-01]
25
-------
Table 5. continued
Variable
Barium
Cadmium
Chromium
Cyanide
Fluoride
Mercury (inorganic)
Nitrate
Nitrite
Selenium
Antimony
Endrin
Metropolitan-Urbanized
(RUCC1 = 1,167)
Units Mean (SD) [Range]
8.08E-02 (3.93E-01) [1.00E-02,
mg/L 1.31E+01]
1.71E-03 (8.60E-04) [1.00E-03,
mg/L 6.00E-03]
6.21 E-03 (7.16E-03) [1.00E-03,
mg/L 1.46E-01]
1.51E-02 (2.85E-02) [1.00E-03,
mg/L 2.67E-01]
1.16E+00 (7.81 E+00) [2.00E-02,
mg/L 1.50E+02]
1.15E-03 (1.13E-03) [1.00E-03,
mg/L 3.60E-02]
8.07E-01 (1.64E+00) [1.00E-02,
mg/L 2.00E+01]
6.78E-02 (1.76E-01) [1.00E-02,
mg/L 3.60E+00]
4.19E-03 (5.46E-03) [1.00E-03,
mg/L 9.50E-02]
2.51 E-03 (1.76E-03) [1.00E-03,
mg/L 2.00E-02]
8.05E-02 (2.01E-01) [1.00E-02,
mg/L 1.01 E+00]
Nonmetropolitan-
Urbanized (RUCC2
= 306)
Mean (SD) [Range]
8.34E-02 (2.37E-01)
[1.00E-02, 3.98E+00]
1.66E-03 (7.77E-04)
[1.00E-03, 7.00E-03]
6.09E-03 (5.69E-03)
[1.00E-03, 3.60E-02]
1.68E-02 (2.92E-02)
[1.00E-03, 2.11E-01]
4.31E-01 (4.20E-01)
[2.00E-02, 2.65E+00]
1.08E-03 (2.74E-04)
[1.00E-03, 2.00E-03]
6.59E-01 (1.19E+00)
[1.00E-02,1.46E+01]
6.70E-02 (1.39E-01)
[1.00E-02,1.90E+00]
3.82E-03 (3.48E-03)
[1.00E-03, 3.10E-02]
2.50E-03 (1.59E-03)
[1.00E-03, 7.00E-03]
7.26E-02 (1.84E-01)
[1.00E-02,1.01 E+00]
Less Urbanized
(RUCC3 = 1,026)
Mean (SD)
[Range]
6.81E-02 (9.96E-02)
[1.00E-02,1.03E+00]
1.66E-03 (7.67E-04)
[1.00E-03, 8.00E-03]
6.27E-03 (7.48E-03)
[1.00E-03, 5.60E-02]
1.57E-02 (3.18E-02)
[1.00E-03, 3.39E-01]
4.83E-01 (6.44E-01)
[2.00E-02, 8.71 E+00]
1.09E-03 (3.08E-04)
[1.00E-03, 5.00E-03]
7.37E-01 (2.80E+00)
[1.00E-02, 8.10E+01]
5.84E-02 (1.17E-01)
[1.00E-02,1.54E+00]
3.96E-03 (4.21 E-03)
[1.00E-03, 3.10E-02]
2.49E-03 (1.63E-03)
[1.00E-03, 7.00E-03]
7.86E-02 (2.03E-01)
[1.00E-02,1.01 E+00]
Thinly
Populated
(RUCC4 = 644)
Mean (SD)
[Range]
4.84E-02 (7.72E-
02) [1.00E-02,
6.70E-01]
1.45E-03 (6.96E-
04) [1.00E-03,
7.00E-03]
4.21 E-03 (6.36E-
03) [1.00E-03,
1.01E-01]
1.39E-02 (4.12E-
02) [1.00E-03,
8.16E-01]
3.50E-01 (6.63E-
01) [2.00E-02,
1.14E+01]
1.08E-03 (3.44E-
04) [1.00E-03,
7.00E-03]
6.22E-01
(2.01 E+00) [1.00E-
02, 3.28E+01]
5.18E-02 (1.71E-
01) [1.00E-02,
3.41 E+00]
3.21 E-03 (4.50E-
03) [1.00E-03,
4.80E-02]
2.00E-03 (1.44E-
03) [1.00E-03,
7.00E-03]
5.71 E-02 (1.75E-
01) [1.00E-02,
1.01E+001
Total (3143)
Mean (SD)
[Range]
7.03E-02 (2.59E-
01) [1.00E-02,
1.31E+01]
1.64E-03 (7.96E-
04) [1.00E-03,
8.00E-03]
5.81 E-03 (7.02E-
03) [1.00E-03,
1.46E-01]
1.52E-02 (3.26E-
02) [1.00E-03,
8.16E-01]
7.02E-01
(4.80E+00)
[2.00E-02,
1.50E+02]
1.11 E-03 (7.33E-
04) [1.00E-03,
3.60E-02]
7.32E-01
(2.13E+00)
[1.00E-02,
8.10E+01 ]
6.13E-02 (1.55E-
01) [1.00E-02,
3.60E+00]
3.88E-03 (4.72E-
03) [1.00E-03,
9.50E-02]
2.40E-03 (1.65E-
03) [1.00E-03,
2.00E-02]
7.43E-02 (1.95E-
01) [1.00E-02,
1.01 E+001
4.76E-01
1.59E-01 (6.02E- (1.52E+00)
6.90E-01 (1.97E+00) [1.00E-02, 5.66E-01 (1.65E+00) 4.06E-01 (1.23E+00) 01) [1.00E-02, [1.00E-02,
Methoxychlor pg/L 1.00E+01] [1.00E-02,1.00E+01] [1.00E-02,9.65E+00] 8.01E+00] 1.00E+01]
7.89E+00
7.78E+00 (2.41E+01)
7.28E+00 (2.27E+01) [8.00E-02, 8.47E+00 (2.44E+01) 8.47E+00 (2.50E+01) (2.49E+01) [8.00E- [8.00E-02,
Dalapon |jg/L 1.00E+02] [8.00E-02,1.00E+02] [8.00E-02,1.00E+02] 02,1.00E+02] 1.00E+02]
5.74E+00
1.30E+00 (1.79E+02)
1.12E+01 (2.94E+02) [6.00E-02, 3.15E+00 (9.77E+00) 3.03E+00 (1.77E+01) (5.69E+00) [6.00E- [6.00E-02,
Di(2-ethyIhexyl) adipate pg/L 1.00E+04] [6.00E-02,5.01E+01] [6.00E-02,5.01E+02] 02,5.01E+01] 1.00E+04]
1.56E-01 (2.31E- 2.10E-01 (2.94E-
2.25E-01 (3.12E-01) [5.00E-02, 2.38E-01 (3.77E-01) 2.19E-01 (2.77E-01) 01) [5.00E-02, 01) [5.00E-02,
Simazine pg/L 4.89E+00] [5.00E-02,5.05E+00] [5.00E-02,1.85E+00] 1.05E+00] 5.05E+00]
7.57E-01
4.79E-01 (8.87E- (1.21 E+00)
8.55E-01 (1.26E+00) [8.00E-02, 8.72E-01 (1.20E+00) 7.87E-01 (1.29E+00) 01) [8.00E-02, [8.00E-02,
Di(2-ethyIhexyl) pthalate pg/L 9.41E+00] [8.00E-02,6.08E+00] [8.00E-02,1.59E+01] 9.15E+00] 1.59E+01]
2.34E+00
1.22E+00 (9.71 E+00)
2.44E+00 (1.00E+01) [4.00E-02, 3.64E+00 (1.25E+01) 2.54E+00 (1.00E+01) (6.36E+00) [4.00E- [4.00E-02,
Picloram pg/L 5.00E+01] [4.00E-02,1.00E+02] [4.00E-02,5.00E+01] 02,5.00E+01] 1.00E+02]
26
-------
Table 5. continued
Variable
Units
Metropolitan-Urbanized
(RUCC1 = 1,167)
Mean (SD) [Range]
Nonmetropolitan-
Urbanized (RUCC2
= 306)
Mean (SD) [Range]
Less Urbanized
(RUCC3 = 1,026)
Mean (SD)
[Range]
Thinly
Populated
(RUCC4 = 644)
Mean (SD)
[Range]
Total (3143)
Mean (SD)
[Range]
Dinoseb
pg/L
2.94E-01 (4.19E-01) [8.00E-02,
3.08E+00]
3.32E-01 (4.45E-01)
[8.00E-02, 2.08E+00]
2.92E-01 (4.64E-01)
[8.00E-02, 9.08E+00]
2.48E-01 (3.87E-
01) [8.00E-02,
2.08E+00]
2.88E-01 (4.31 E-
01) [8.00E-02,
9.08E+00]
Atrazine
pg/L
2.05E-01 (3.12E-01) [3.00E-02,
2.53E+00]
2.24E-01 (3.42E-01)
[3.00E-02, 3.78E+00]
2.73E-01 (2.37E+00)
[3.00E-02, 7.53E+01]
1.34E-01 (2.35E-
01) [3.00E-02,
2.28E+00]
2.15E-01
(1.37E+00)
[3.00E-02,
7.53E+01]
2,4-Dichlorophenoxyacetic acid
Mg/L
1.40E-01 (1.08E-01) [9.00E-02,
2.51 E+00]
1.42E-01 (5.41 E-02)
[9.00E-02, 4.00E-01]
1.42E-01 (2.27E-01)
[9.00E-02, 7.19E+00]
1.20E-01 (5.30E-
02) [9.00E-02,
8.10E-01]
1.37E-01 (1.49E-
01) [9.00E-02,
7.19E+00]
Benzo[a]pyrene
pg/L
4.78E-02 (5.40E-02) [1.00E-02,
3.47E-01]
5.03E-02 (5.82E-02)
[1.00E-02, 3.34E-01]
5.33E-02 (5.93E-02)
[1 00E-02, 3.10E-01]
3.84E-02 (4.93E-
02) [1.00E-02,
2.10E-01]
4.79E-02 (5.56E-
02) [1.00E-02,
3.47E-01]
Pentachlorophenol
Mg/L
7.84E-02 (1.63E-01) [1.00E-02,
171 E+00]
8.91E-02 (1.81E-01)
[1 00E-02, 1.01 E+00]
8.82E-02 (1.76E-01)
[1.00E-02,1.01 E+00]
6.16E-02 (1.36E-
01) [1 00E-02,
1.01 E+00]
7.92E-02 (1.65E-
01) [1.00E-02,
1.71 E+00]
Polychlorinated biphenyls
pg/L
1.65E-01 (1.19E+00) [6.00E-02,
4.04E+01]
1.13E-01 (1.24E-01)
[6.00E-02,1.06E+00]
1.13E-01 (1.88E-01)
[6.00E-02, 4.31 E+00]
8.13E-02 (6.53E-
02) [6.00E-02,
1.06E+00]
1.26E-01 (7.35E-
01) [6.00E-02,
4.04E+01]
1,2-Dibromo-3-chloropropane
Mg/L
2.19E-02 (1.93E-02) [1.00E-02,
5.45E-01]
2.01 E-02 (9.92E-03)
[1 00E-02, 3.00E-02]
2.05E-02 (9.96E-03)
[1 00E-02, 4.50E-02]
1.86E-02 (9.86E-
03) [1 00E-02,
3.00E-02]
2.06E-02 (1.42E-
02) [1.00E-02,
5.45E-01]
Ethylene dibromide
pg/L
8.28E-02 (1.60E-01) [1.00E-02,
1.17E+00]
7.14E-02 (1.39E-01)
[1.00E-02, 5.10E-01]
6.94E-02 (1.41 E-01)
[1 00E-02, 8.70E-01]
8.19E-02 (1.59E-
01) [1.00E-02,
5.10E-01]
7.72E-02 (1.52E-
01) [1.00E-02,
1.17E+00]
Xylenes
Mg/L
8.44E-01 (6.05E+00) [1.00E-01,
2.00E+02]
8.60E-01 (3.26E+00)
[1.00E-01, 5.08E+01]
2.00E+00 (4.37E+01)
[1.00E-01,1.40E+03]
2.01 E+00
(3.94E+01) [1.00E-
01, 1.00E+03]
1.46E+00
(3.09E+01)
[1.00E-01,
1.40E+03]
Chlordane
pg/L
1.08E-01 (9.94E-02) [2.00E-02,
9.70E-01]
1.17E-01 (9.62E-02)
[2.00E-02, 2.76E-01]
1.12E-01 (9.77E-02)
[2.00E-02, 2.87E-01]
8.43E-02 (9.23E-
02) [2.00E-02,
2.20E-01]
1.06E-01 (9.77E-
02) [2.00E-02,
9.70E-01]
Dichloromethane
pg/L
4.99E-01 (4.91E-01) [1.00E-01,
1.03E+01]
4.90E-01 (2.67E-01)
[1.00E-01,1.98E+00]
4.95E-01 (3.09E-01)
[1.00E-01, 4.05E+00]
4.29E-01 (5.13E-
01) [1 00E-01,
1.18E+01]
4.83E-01 (4.27E-
01) [1.00E-01,
1.18E+01]
p-Dichlorobenzene
pg/L
5.09E-01 (5.13E+00) [2.00E-02,
175E+02]
3.72E-01 (2.41 E-01)
[2.00E-02,1.54E+00]
3.62E-01 (2.57E-01)
[2.00E-02, 2.77E+00]
3.11 E-01 (3.55E-
01) [2.00E-02,
6.02E+00]
4.07E-01
(3.13E+00)
[2.00E-02,
1.75E+02]
1,1,1 -T richloroethane
Mg/L
6.77E-01 (1.03E+01) [1.00E-02,
3.51 E+02]
7.94E-01 (7.15E+00)
[1 00E-02, 1.25E+02]
3.99E-01 (9.67E-01)
[1 00E-02, 3.03E+01]
3.03E-01 (2.51 E-
01) [1 00E-02,
2.16E+00]
5.21E-01
(6.67E+00)
[1.00E-02,
3.51 E+02]
Trichloroethylene
pg/L
4.39E-01 (4.89E-01) [2.00E-02,
6.50E+00]
4.06E-01 (2.67E-01)
[2.00E-02, 2.03E+00]
4.00E-01 (2.70E-01)
[2.00E-02, 3.75E+00]
3.27E-01 (2.54E-
01) [2.00E-02,
1.93E+00]
4.00E-01 (3.67E-
01) [2.00E-02,
6.50E+00]
Carbon tetrachloride
Mg/L
4.62E-01 (5 79E-01) [1 OOE-02,
8.01 E+00]
4.13E-01 (3.76E-01)
[1 00E-02, 5.12E+00]
4.22E-01 (7.75E-01)
[1.00E-02, 2.38E+01]
3.26E-01 (2.96E-
01) [1 00E-02,
4.34E+00]
4.16E-01 (5.95E-
01) [1.00E-02,
2.38E+01]
Benzene
pg/L
4.92E-01 (3.48E-01) [1.10E-01,
4.24E+00]
4.87E-01 (2.43E-01)
[1.10E-01,1.74E+00]
4.94E-01 (2.47E-01)
[1.10E-01, 3.24E+00]
4.22E-01 (2.49E-
01) [1.10E-01,
1.55E+00]
4.78E-01 (2.90E-
01) [1.10E-01,
4.24E+00]
Toluene
Mg/L
7.60E-01 (6.22E+00) [7.00E-02,
2.01 E+02]
2.59E+00 (2.27E+01)
[7.00E-02, 3.34E+02]
1.07E+00 (1.26E+01)
[7.00E-02, 3.50E+02]
4.43E-01
(1.34E+00) [7.00E-
02, 3.37E+01]
9.74E-01
(1.08E+01)
[7.00E-02,
3.50E+02]
Ethylbenzene
pg/L
5.00E-02 (0.00E+00) [5.00E-02,
5.00E-02]
5.00E-02 (0.00E+00)
[5.00E-02, 5.00E-02]
5.00E-02 (0.00E+00)
[5.00E-02, 5.00E-02]
5.00E-02
(0.00E+00) [5.00E-
02, 5.00E-02]
5.00E-02
(0.00E+00)
[5.00E-02,
5.00E-02]
27
-------
Table 5. continued
Variable
Units
Metropolitan-Urbanized
(RUCC1 = 1,167)
Mean (SD) [Range]
Nonmetropolitan-
Urbanized (RUCC2
= 306)
Mean (SD) [Range]
Less Urbanized
(RUCC3 = 1,026)
Mean (SD)
[Range]
Thinly
Populated
(RUCC4 = 644)
Mean (SD)
[Range]
Total (3143)
Mean (SD)
[Range]
Styrene
pg/L
5.67E-01 (2.37E+00) [1.00E-01,
7.86E+01]
4.91 E-01 (3.40E-01)
[1.00E-01, 3.58E+00]
4.93E-01 (3.12E-01)
[1.00E-01, 5.00E+00]
4.14E-01 (2.73E-
01) [1.00E-01,
2.80E+00]
5.04E-01
(1 47E+00)
[1.00E-01,
7.86E+01]
Alpha particles
pCi/L
1.05E+00 (2.32E+00) [0.00E+00,
3.58E+01]
1.24E+00 (3.42E+00)
[0.00E+00, 5.15E+01]
1.34E+00 (3.19E+00)
[0.00E+00, 3.47E+01]
7.33E-01
(2.01 E+00)
[0.00E+00,
1.81E+01]
1.10E+00
(2.71 E+00)
[0.00E+00,
5.15E+01]
cis-1,2-Dichloroethylene
pg/L
3.87E-01 (4 21E-01) [2.00E-02,
1.19E+01]
4.02E-01 (3.53E-01)
[2.00E-02, 5.19E+00]
3.92E-01 (2.22E-01)
[2.00E-02,1.22E+00]
3.28E-01 (2.53E-
01) [2.00E-02,
2.09E+00]
3.78E-01 (3.28E-
01) [2.00E-02,
1.19E+01]
Construct: Drinking Water Quality
Total coliform proportion
Proportion
1.20E-01 (3.55E-01) [1.00E-03,
4.93E+00]
2.86E-01 (1.26E+00)
[1.00E-03, 1.34E+01]
2.03E-01 (8.82E-01)
[1.00E-03,1.84E+01]
2.22E-01 (8.41 E-
01) [1 00E-03,
9.71 E+00]
1.84E-01 (7.76E-
01) [1.00E-03,
1.84E+01]
Land Domain
In the land domain, the metropolitan-urbanized counties had
lower agricultural-related variables (percent harvested and
percent irrigated) than did nonmetropolitan-urbanized, less urban,
and thinly populated counties (Table 6). Pesticides and animal
units showed no clear pattern in variation across the strata. For
example, average pounds of herbicides applied were 58,700,
78,400, 75,100, and 61,500 for most urban to most rural strata,
respectively. There was little variation in the distribution of radon
zones across the urban/rural strata.
Table 6. Land domain variable means, standard deviations (SDs), and ranges - Overall and rural-urban continuum codes (RUCCs)
stratified
Variable
Construct: Agriculture
Farms per acre
Irrigated acreage
Chemicals used to control
nematodes, acres applied per
county acres
Manure, acres applied per
county acres
Chemicals used to control
diseases in crops and
orchards, acres applied per
county acres
Chemicals used to defoliate/
control growth/thin fruit, acres
applied per county acres
Harvested acreage, acres
harvested per county acres
Animal Units, animal units per
county acres
Metropolitan-
Urbanized (RUCC1
= 1,167) Mean (SD)
Units [Range]
1.53E-03 (1.10E-03)
Number [2.34E-06,7.87E-03]
2.20E+00 (6.72E+00)
% [3.62E-04,7.42E+01]
1.01E-02 (1.28E-02)
% [1.32E-06,1.07E-01]
1.69E-02 (2.56E-02)
% [1.56E-06,2.63E-01]
1 48E-02 (2.62E-02)
[8.78E-07, 2.25E-01]
1.46E-02 (2.91E-02)
[8.49E-07, 3.84E-01]
1.90E-01 (2.12E-01)
[2.59E-05, 9.94E-01]
2.62E-04 (1.01E-03)
[1.31E-08,1.75E-02]
Nonmetropolitan-
Urbanized (RUCC2
= 306) Mean (SD)
[Range]
1 49E-03 (1 06E-03) [2.34E-
06, 6.48E-03]
3.46E+00 (9.15E+00) [3.62E-
04, 5.65E+01]
1.14E-02 (1.54E-02) [1.32E-
06,1.30E-01]
2.10E-02 (2.71 E-02) [1.56E-
06,1.68E-01]
1.68E-02 (2.63E-02) [8.78E-
07,1.59E-01]
1.67E-02 (3.28E-02) [8.49E-
07, 3.63E-01]
2.47E-01 (2.50E-01) [2.59E-
05, 9.16E-01]
1.11E-04 (2.08E-04) [1.31E-
08, 2.36E-03]
Less Urbanized
(RUCC3 = 1026)
Mean (SD) [Range]
1.34E-03 (1.03E-03)
[2.34E-06, 5.95E-03]
3.45E+00 (8.73E+00)
[3.62E-04, 7.14E+01]
1.27E-02 (1.60E-02)
[1.32E-06, 1.50E-01]
1.96E-02 (2.83E-02)
[1.56E-06, 2.52E-01]
1.86E-02 (3.06E-02)
[8.78E-07, 2.60E-01]
1.91 E-02 (3.37E-02)
[8.49E-07, 4.15E-01 ]
2.51E-01 (2.60E-01)
[2.59E-05, 9.43E-01]
1.29E-04 (4.09E-04)
[1.31E-08, 6.14E-03]
Thinly Populated
(RUCC4 = 644)
Mean (SD)
[Range]
9.15E-04 (8.72E-04)
[2.34E-06, 5.18E-03]
2.81 E+00 (7.39E+00)
[3.62E-04, 6.07E+01]
8.75E-03 (1 08E-02)
[1.32E-06, 9.63E-02]
1.12E-02 (1.78E-02)
[1.56E-06,1.54E-01]
1.95E-02 (3.32E-02)
[8.78E-07, 3.05E-01]
1.32E-02 (1.92E-02)
[8.49E-07, 2.12E-01]
2.18E-01 (2.25E-01)
[2.59E-05, 9.21E-01]
1.32E-04 (5.43E-04)
[1.31 E-08, 6.75E-03]
Total (3143)
Mean (SD)
[Range]
1.34E-03 (1.05E-
03) [2.34E-06,
7.87E-03]
2.86E+00
(7.83E+00) [3.62E-
04, 7.42E+01]
1.08E-02 (1.39E-
02) [1.32E-06,
1.50E-01]
1.70E-02 (2.55E-
02) [1.56E-06,
2.63E-01]
1.72E-02 (2.93E-
02) [8.78E-07,
3.05E-01]
1.60E-02 (2.95E-
02) [8.49E-07,
4.15E-01]
2.21E-01 (2.37E-
01) [2.59E-05,
9.94E-01]
1.77E-04 (7.11 E-
04) [1.31 E-08,
1.75E-02]
28
-------
Table 6. continued
Variable
Units
Metropolitan-
Urbanized (RUCC1
= 1,167) Mean (SD)
[Range]
Nonmetropolitan-
Urbanized (RUCC2
= 306) Mean (SD)
[Range]
Less Urbanized
(RUCC3 = 1026)
Mean (SD) [Range]
Thinly Populated
(RUCC4 = 644)
Mean (SD)
[Range]
Total (3143)
Mean (SD)
[Range]
Construct: Pesticides
Fungicides, applied
Pounds
2.66E+04 (2.00E+05)
[3.75E-01, 5.17E+06]
8.56E+03 (2.44E+04) [3.00E-
01, 2.24E+05]
6.37E+03 (1.74E+04)
[2.00E-01, 2.37E+05]
3.96E+03 (9.61 E+03)
[4.33E-01, 1.59E+05]
1.36E+04
(1.23E+05) [2.00E-
01, 5.17E+06]
Herbicides, applied
Pounds
5.87E+04 (8.30E+04)
[2.23E+00, 8.68E+05]
7.84E+04 (9.32E+04) [7.00E-
01, 6.17E+05]
7.51 E+04 (8.39E+04)
[1.42E+01, 4.75E+05]
6.15E+04 (7.00E+04)
[2.00E-01, 4.28E+05]
6.65E+04
(8.22E+04) [2.00E-
01,8.68E+05]
Insecticides, applied
Pounds
9.61 E+03 (3.23E+04)
[2.00E-01, 5.72E+05]
8.96E+03 (2.11 E+04)
[2.01 E+01, 2.30E+05]
8.11 E+03 (1.42E+04)
[1.85E+00, 2.57E+05]
5.18E+03 (7.47E+03)
[1.00E-01, 9.77E+04]
8.15E+03
(2.26E+04) [1.00E-
01,5.72E+05]
Construct: Mines
Primarily coal mines, mines
per county pop.
Proportion
1.11E-04 (7.38E-04)
[6.25E-07,1.25E-02]
1.35E-04 (5.64E-04) [6.25E-
07, 4.67E-03]
4.05E-04 (2.18E-03)
[6.25E-07, 2.82E-02]
5.67E-04 (3.75E-03)
[6.25E-07, 5.78E-02]
3.03E-04 (2.17E-
03) [6.25E-07,
5.78E-02]
Primarily metal mines, mines
per county pop.
Proportion
3.29E-05 (3.24E-04)
[2.44E-07, 6.43E-03]
4.14E-05 (2.19E-04) [2.44E-
07, 2.54E-03]
1.19E-04 (7.78E-04)
[2.44E-07, 1.43E-02]
5.18E-04 (3.84E-03)
[2.44E-07, 7.41 E-02]
1.61E-04 (1.81E-
03) [2.44E-07,
7.41 E-02]
Primarily nonmetal mines,
mines per county pop.
Proportion
3.16E-05 (2.57E-04)
[2.86E-07, 7.67E-03]
3.08E-05 (7.09E-05) [2.86E-
07, 6.35E-04]
7.76E-05 (3.34E-04)
[2.86E-07, 6.41E-03]
1.43E-04 (8.15E-04)
[2.86E-07,1.66E-02]
6.94E-05 (4.46E-
04) [2.86E-07,
1.66E-02]
Primarily sand and gravel
mines, mines per county pop.
Proportion
1.40E-04 (3.49E-04)
[2.00E-07, 6.87E-03]
2.07E-04 (2.38E-04) [2.00E-
07,1.25E-03]
3.47E-04 (4.78E-04)
[2.00E-07, 4.43E-03]
8.32E-04 (1.34E-03)
[2.00E-07,1.24E-02]
3.56E-04 (7.49E-
04) [2.00E-07,
1.24E-02]
Primarily stone mines, mines
per county pop.
Proportion
9.42E-05 (3.10E-04)
[3.06E-07, 5.66E-03]
1.12E-04 (1.78E-04) [3.06E-
07,1.95E-03]
2.04E-04 (5.12E-04)
[3.06E-07, 9.32E-03]
3.40E-04 (1.32E-03)
[3.06E-07, 2.42E-02]
1.82E-04 (7.00E-
04) [3.06E-07,
2.42E-02]
Construct: Radon
Radon
Ordinal
2.02E+00 (8.14E-01)
[0.00E+00, 3.00E+00]
1.97E+00 (8.23E-01)
[1.00E+00, 3.00E+00]
2.03E+00 (8.24E-01)
[1.00E+00, 3.00E+00]
1.88E+00 (8.09E-01)
[1.00E+00, 3.00E+00]
1.99E+00 (8.19E-
01) [0.00E+00,
3.00E+00]
Construct: Facilities
Facilities per county
Proportion
3.69E-04 (2.82E-04)
[5.60E-06, 3.22E-03]
4.99E-04 (3.25E-04) [3.69E-
05, 2.24E-03]
5.60E-04 (4.63E-04)
[5.60E-06, 6.65E-03]
8.25E-04 (2.08E-03)
[5.60E-06, 4.58E-02]
5.38E-04 (1.01E-
03) [5.60E-06,
4.58E-02]
Sociodemographic Domain lowest household income ($30,300) and lowest household value
Socioeconomic variables included in the sociodemographic ($94,900). From the crime perspective, however, rural areas
domain indicated that rural counties generally were more were at an advantage compared with more urban areas; the mean
deprived than were more urban counties (Table 7), with both the violent crime rate per county population for rural counties was
385.5 compared with 619.8 for the most urban counties.
Table 7. Sociodemographic domain variable means, standard deviations (SDs), and ranges - Overall and rural-continuum codes (RUCCs)
stratified
Metropolitan-
Urbanized
(RUCC1 = 1167)
Mean (SD)
Variable Units [Range]
Sociodemographic Domain
Construct: Socioeconomic
15.1(5.8) 12.7(4.6) 10.5(4.0) 11.4(4.6) 12.6(5.3)
Percent bachelor's degree % [2.6,37.2] [5.4,34.7] [3.0,42.2] [1.9,36.1] [1.9,42.2]
7.6(2.5) 8.1(2.6) 7.9(3.4) 6.7(4.6) 7.5(3.6)
Percent unemployed % [0,27.5] [2.2,20.2] [0.3,26.3] [0.0,30.9] [0,30.9]
Percent families less than 9.8(4.5) 11.9(4.8) 12.7(5.8) 11.9(6.4) 11.4(5.5)
poverty level % [0,39.6] [3.1,35.1] [1.4,44.9] [0.0,44.4] [0,44.9]
Nonmetropolitan-
Urbanized
(RUCC2 = 306)
Mean (SD)
[Range]
Less
Urbanized
(RUCC3 = 1026)
Mean (SD)
[Range]
Thinly
Populated
(RUCC4 = 644)
Mean (SD)
[Range]
OVERALL
(n=3143)
Mean (SD)
[Range]
29
-------
Table 7. continued
Metropolitan-
Nonmetropolitan-
Less
Thinly
Urbanized
Urbanized
Urbanized
Populated
OVERALL
(RUCC1 = 1167)
(RUCC2 = 306)
(RUCC3 = 1026)
(RUCC4 = 644)
(n=3143)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
Mean (SD)
Variable
Units
[Range]
[Range]
[Range]
[Range]
[Range]
12.1 (6.5)
14.8 (7.7)
18.5(9.3)
25.8(12.3)
17.3(10.3)
Percent vacant housing
%
[1.7,60.1]
[5.5, 63.9]
[4.9, 68.0]
[7.2, 83.3]
[1.7,83.3]
175.4(103.9)
135.4(78.7)
106.6 (64.9)
94.9 (55.5)
133.5 (88.4)
Median household value (X1000)
Dollar value
[0, 868k]
[57.0, 583.2k]
[18.6,100.0k]
[29.7, 4965.6k]
[0,1000k]
82.6(17.0)
23.1 (9.7)
8.7 (4.9)
3.0 (2.4)
36.3k (109.9)
Household income (X1000)
Dollars
[67.0,3217.9k]
[5.9, 76.7k]
[1.1,30.7k]
[0.2,15.4k]
[22,321.8k]
0.6 (0.6)
0.6 (0.6)
0.8(1.2)
0.9(1.4)
0.7(1.0)
Count of occupants per room
Count
[0.1,6.1]
[0.1,5.4]
[0.1,20.2]
[0.1,31.5]
[0.1,31.5]
28.0 (9.3)
30.0 (6.3)
26.2 (5.9)
23.6 (7.0)
26.7 (7.8)
Percent renter-occupied housing
%
[8.7,100]
[16.8, 51.0]
[11.3,53.7]
[8.7,71.4]
[8.7,100]
0.43 (0.04)
0.44 (0.03)
0.4(0.0)
0.4(0.0)
0.43 (0.04)
Gini coefficient
Proportion
[0.3, 0.6]
[0.35, 0.54]
[0.3, 0.6]
[0.2, 0.6]
[0.21,0.65]
Construct: Crime
Rate per
Mean number of violent crimes
county
619.8 (441.4)
472.3 (308.2)
446.7 (249.8)
385.5 (195.1)
500.9 (344.5)
per capita
population
[22.6, 6628.6]
[19.52,1735.0]
[7.3,1710.7]
[69.9,1420.1]
[7.3, 6628.6]
Construct: County typology
0.2 (0.1)
0.2 (0.0)
0.2 (0.0)
0.15(0.0)
0.18(0.06)
Creative class
%
[0,0.51]
[0.1,0.4]
[0.0, 0.5]
[0, 0.4]
[0,0.51]
Construct: County political
valence
44.8(13.7)
43.9(12.4)
40.2(12.9)
36.4(14.3)
41.5(13.8)
Percent Democratic voters
%
[5.5, 92.5]
[12.5,84.5]
[7.8, 88.7]
[4.9, 86.8]
[4.9, 92.5]
NOTE: Means calculated using nontransformed variables
k = 1000
Built Domain
The most urban counties had a higher rate of traffic fatalities and
residents reporting spending more time commuting compared
with more rural areas (Table 8). Urban counties also had a higher
walkability score but contained less green space and undeveloped
areas than rural counties.
Table 8. Built-environment domain variable means, standard deviations (SDs), and ranges - Overall and rural-urban continuum codes
(RUCCs) stratified
Metropolitan-
Urbanized
(RUCC1 = 1167)
Mean (SD)
Variable Units [Range]
Built Domain
Construct: Business environment
Vice-related environment
Count/county
population
4.9e-4 (3.1 e-4)
[1,5e-5, 3.4e-3]
5.8e-4 (2.9e-4)
[6.3e-5,1.8e-3]
6.4e-4 (4.3e-4)
[1.5e-5, 2.8e-3]
8.9e-4 (8.9e-3)
[1.5e-5, 7.2e-3]
6.3e-4 (5.3e-4)
[1.5e-5, 7.2e-3]
Civic-related environment
Count/ county
population
2.9e-3 (9.4e-4)
[2.5e-4, 8.4e-4]
3.3e-3 (8.6e-4)
[9.5e-4, 7.2e-3]
3.8e-3 (1.1 e-3)
[5.9e-4, 6.5e-3]
4.3e-3 (1.7e-3)
[2.5e-4,1,6e-2]
3.5e-3 (1.3e-3)
[2.5e-4,1,6e-2]
Education-related environment
Count/ county
population
1.2e-3 (4.2e-4)
[1,8e-4, 4.5e-3]
1.3e-3 (3.6e-4)
[6.3e-4, 3.2e-3]
1,5e-3 (6.0e-4)
[5.9e-4, 6.5e-3]
2.5e-3 (1.8e-3)
[1.8e-4,1.8e-2]
1.6e-3 (1 Oe-3)
[1.8e-4,1.8e-2]
Health care-related environment
Count/ county
population
3.4e-3(1.6e-3)
[3.4e-3,1.6e-3]
3.7e-3 (1.1 e-3)
[1.0e-3,1.1 e-2]
3.2e-3 (1.3e-3)
[6.0e-4, 2.0e-2]
2.8e-3 (1.4e-3)
[1 Oe-4, 9.1 e-3]
3.2e-3 (1.4e-3)
[1 Oe-4, 2.0e-2]
Negative food environment
Count/ county
population
1.2e-3 (3.4e-4)
[7.0e-5, 3.4e-3]
1.4e-3 (3.8e-4)
[6.4e-4, 4.3e-3]
1.4e-3 (4.2e-4)
[1.7e-4, 4.7e-3]
1,3e-3 (8.5e-4)
[7.0e-5, 1.3e-2]
1,3e-3 (5.2e-4)
[7.0e-5,1.3e-2]
Positive food environment
Count/ county
population
2.2e-3 (7.7e-7)
[1,3e-4,8.1e-3]
2.3e-3 (8.5e-4)
[1 Oe-3, 7.8e-3]
2.4e-3 (8.9e-4)
[4.4e-4, 9.0e-3]
2.9e-3 (1.7e-3)
[1.3e-4, 2.0e-2]
2.4e-3 (1.1e-3)
[1,3e-4, 2.0e-2]
Nonmetropolitan-
Urbanized
(RUCC2 = 306)
Mean (SD)
[Range]
Less
Urbanized
(RUCC3 = 1026)
Mean (SD)
[Range]
Thinly
Populated
(RUCC4 = 644)
Mean (SD)
[Range]
OVERALL
(n=3143)
Mean (SD)
[Range]
30
-------
Table 8. continued
Variable
Recreation environment
Social service-related environment
Construct: Highway safety
Traffic fatality rate
Construct: Housing
Rate of low-rent + Section 8
housing
Construct: Roads
Proportion of roads that are
secondary
Construct: Commuting practices
Residents who report using public
transport
Commute time
Construct: Walkability
Walkability score
Construct: Green space
County land area classified as
natural cover and open space %
NOTE: Means calculated using nontransformed variable
Units
Count/county
population
Count/ county
population
Fatality count/
county population
Unit count/ county
population
Secondary road
mile / total road
miles
Minutes
Ordinal
Metropolitan-
Urbanized
(RUCC1 = 1167)
Mean (SD)
[Range]
1.3e-3 (6.1 e-4)
[4.7e-5,1.1 e-2]
1,5e-3 (5.9e-4)
[9.2e-5, 5.1e-3]
23.2 (39.0)
[1.0,685.8]
0.2 (0.4)
[0.0,1.0]
0.2 (0.1)
[0.0, 0.5]
1.8 (4.6)
[0.1,60.5]
25.0(5.1)
[6.2, 60.5]
7.1 (2.3)
[1.7,16.2]
61.5 (24.4)
[3.9, 99.7]
Nonmetropolitan-
Urbanized
(RUCC2 = 306)
Mean (SD)
[Range]
1,6e-3 (8.5e-4)
[3.0e-4, 8.8e-3]
1,8e-3 (6.5e-4)
[6.2e-4, 4.8e-3]
11.2(6.5)
[1.3,59.6]
0.2 (0.4)
[0.0, 1.0]
0.1 (0.1)
[0.0, 0.44]
0.7(1.2)
[0.1,12.8]
20.7 (3.6)
[12.3,31.8]
6.6(1.1)
[4.1,13.8]
62.3 (28.0)
[5.3, 99.8]
Less
Urbanized
(RUCC3 = 1026)
Mean (SD)
[Range]
1.7e-3(1.0e-3)
[1.2e-4,1.0e-2]
1,8e-3 (7.8e-4)
[3.0e-4, 5.2e-3]
5.7 (3.5)
[1.0,39.4]
0.4(0.5)
[0.0,1.0]
0.14(0.1)
[0.0, 0.4]
0.7(1.0)
[0.1,13.0]
21.6 (5.0)
[5.4, 38.5]
5.9(1.1)
[2.0,10.5]
63.2 (28.6)
[6.9,100.0]
Thinly
Populated
(RUCC4 = 644)
Mean (SD)
[Range]
2.2e-3 (1.9e-3)
[4.7e-5,1,8e-2]
1.9e-3 (1.1 e-3)
[9.2e-5, 8.4e-3]
2.8(1.8)
[1.0,14.0]
0.6 (0.5)
[0.0,1.0]
0.1 (0.1)
[0.1,24.1]
0.9(1.2)
[0.1,24.1]
21.4(6.4)
[4.3, 44.2]
5.3(1.2)
[1.0,9.5]
68.5 (27.7)
[6.2,100.0]
OVERALL
(n=3143)
Mean (SD)
[Range]
1.6e-3 (1.2e-3)
[4.7e-5,1.8e-2]
1.7e-3 (8.0e-4)
[9.2e-5, 8.4e-3]
12.1 (25.5)
[1,685.8]
0.4(0.5)
[0,1]
(0.1)
[0.2, 0.5]
1.2 (3.0)
[0.1,60.5]
22.7 (5.5)
[4.3, 44.2]
6.3(1.8)
[1.0,16.2]
63.5 (27.0)
[3.9,100.0]
Variable Loadings on Environmental Quality Index Domains
Air Domain
The loadings for the variables comprising the air domain are
displayed in Table 9. Each variable lias been annotated with a "+"
or "-"that is the predicted direction for the loading. Because we
want to ensure that higher values of the EQI are associated with
worse enviromnental quality, those variables that we anticipate
being associated with poor enviromnental quality are assigned
a "+" indicating more of this attribute would be a negative for
health. All variables except for S02 and benzidine (in certain
strata) loaded as intended; loadings for S02 and benzidine were
relatively low. Most variables loaded consistently across rural-
urban strata.
Table 9. Variable loadings, valence determination of variables - Air domain
1,1,2-Trichloroethane (
0.0007
0.0016
0.0687
0.0272
0.0224
0.0402
0.0273
0.0728
0.0398
0.1306
0.1665
0.1652
0.1626
0.1514
0.0443
0.0718
0.0794
0.0738
0.0798
-0.0036
-0.0141
-0.0200
-0.0535
-0.0221
0.1345
0.1215
0.1458
0.1745
0.1513
0.1054
0.1191
0.1220
0.1204
0.1278
0.1410
0.1208
0.1478
0.1551
0.1475
0.1120
0.1181
0.1131
0.1042
0.1179
Metropolitan-Urbanized
(RUCC1 = 1167)
Nonmetropolitan-Urbanized
(RUCC2 = 306)
Less Urbanized
(RUCC3 = 1026)
Thinly Populated
(RUCC4 = 644)
OVERALL
(n=3143)
0.1120
0.0443
0.1410
0.1654
0.1181
0.0718
0.1208
0.1508
0.1131
0.0794
0.1478
0.1583
0.1042
0.0738
0.1551
0.1648
0.1179
0.0798
0.1475
0.1616
31
-------
Table 9. continued.
Air Domain
1,2-Dibromo-3-chloropropane
M
1,2-Dichloropropane (+)
Acrylic acid (+)
Benzidine (+)
Benzyl chloride (+)
Beryllium compounds (+)
bis-2-Ethylhexyl phthalate (+)
Carbon tetrachloride (+)
Carbonyl sulfide (+)
Chlorine (+)
Chlorobenzene (+)
Chloroform (+)
Chloroprene (+)
Chromium compounds (+)
Cobalt compounds (+)
Cyanide compounds (+)
Dibutylphthalate (+)
Ethyl chloride (+)
Ethyl benzene (+)
Ethyl dichloride (+)
Glycol ethers (+)
Hydrazine (+)
Hydrochloric acid (+)
Isophorone (+)
Manganese compounds (+)
Methyl bromide (+)
Methyl chloride (+)
Phosphine (+)
Polychlorinated biphenyls (+)
Propylene dichloride (+)
Quinoline (+)
Trichloroethylene (+)
Vinyl chloride (+)
Metropolitan-Urbanized
(RUCC1 = 1167)
0.0722
0.1069
0.1714
-0.0031
0.1976
0.1761
0.1046
0.0649
0.1524
0.1791
0.2065
0.1880
0.1724
0.2012
0.2120
0.1722
0.1923
0.1890
0.2407
0.1275
0.1882
0.1219
0.1910
0.1597
0.1229
0.1404
0.1931
0.0041
0.0971
0.1585
0.1805
0.2283
0.1781
Nonmetropolitan-Urbanized
(RUCC2 = 306)
0.0657
0.1090
0.1785
0.0023
0.1926
0.1460
0.1343
0.1127
0.1322
0.1972
0.1810
0.1674
0.1560
0.2010
0.2223
0.1532
0.2087
0.2047
0.2313
0.1183
0.1987
0.1434
0.1987
0.1775
0.1369
0.0889
0.1905
0.0014
0.1004
0.1529
0.1881
0.2288
0.1577
Less Urbanized
(RUCC3 = 1026)
0.0416
0.1095
0.1727
-0.0058
0.1968
0.1343
0.0872
0.0761
0.1439
0.1877
0.1998
0.1705
0.1479
0.2010
0.2093
0.2033
0.2029
0.1830
0.2343
0.1299
0.1965
0.1261
0.2066
0.1630
0.1358
0.1183
0.1887
0.0054
0.0933
0.1349
0.1915
0.2296
0.1696
Thinly Populated
(RUCC4 = 644)
0.0879
0.1143
0.1422
0.0592
0.1850
0.1688
0.1654
0.1272
0.1664
0.1775
0.1995
0.1713
0.1443
0.1676
0.1908
0.1910
0.1988
0.1946
0.2138
0.1500
0.1673
0.1186
0.1974
0.1667
0.1187
0.1355
0.1756
0.0439
0.1288
0.1254
0.1560
0.1995
0.1767
OVERALL
(n=3143)
0.0688
0.1129
0.1661
0.0135
0.1917
0.1557
0.1192
0.0823
0.1580
0.1866
0.2014
0.1740
0.1537
0.1904
0.2081
0.1825
0.2000
0.1875
0.2306
0.1344
0.1884
0.1246
0.1994
0.1647
0.1250
0.1247
0.1825
0.0089
0.1040
0.1428
0.1799
0.2210
0.1770
Water Domain
The loadings for the variables that comprise the water domain
are displayed in Table 10. Each variable lias been annotated
with a "+" or that is the predicted direction for the loading.
Because we want to ensure that higher values of the EQI are
associated with worse environmental quality, those variables
that we anticipate being associated with poor enviromnental
quality are assigned a "+" indicating more of this attribute would
be a negative for health. The variables in the drought, chemical
contamination and drinking water quality constructs loaded in
the direction intended; however, some of the variables in the
remaining constructs loaded in the opposite direction intended.
Table 10. Variable loadings, valence determination of variables - Water domain
Water Domain
Construct: Domestic Use
Percent of population on self-supply (+)
Percent of public supply population on surface water (
Metropolitan-
Urbanized
(RUCC1 = 1167)
0.0028
0.0197
Nonmetropolitan
Urbanized (RUCC2 =
306)
0.0155
0.0155
Less Urbanized
(RUCC3 = 1026)
0.0203
-0.0004
Thinly Populated
(RUCC4 = 644)
0.0279
0.0251
Total (All:
3143)
0.0096
0.0191
32
-------
Table 10. continued
Water Domain
Metropolitan-
Urbanized
(RUCC1 = 1167)
Nonmetropolitan
Urbanized (RUCC2 =
306)
Less Urbanized
(RUCC3 = 1026)
Thinly Populated
(RUCC4 = 644)
Total (All =
3143)
Construct: Overall Water Quality
Percent of stream length impaired in county (+)
0.0142
-0.0174
-0.0053
0.0160
0.0111
Construct: General Water Contamination
ALL NPDES permits per 1000 km of stream (+)
-0.0161
-0.0415
-0.0225
0.0164
-0.0009
Construct: Atmospheric Deposition
Calcium precipitation weighted mean (+)
0.0378
0.0199
0.0347
-0.0039
0.0206
Potassium precipitation weighted mean (+)
-0.0108
-0.0236
-0.0075
-0.0291
-0.0204
Nitrate precipitation weighted mean (+)
0.0239
0.0014
0.0182
0.0009
0.0140
Chloride precipitation weighted mean (+)
-0.0408
-0.0329
-0.0457
-0.0077
-0.0278
Sulfate precipitation weighted mean (+)
-0.0162
-0.0217
-0.0086
0.0209
-0.0035
Total mercury deposition (+)
-0.0730
-0.0632
-0.0596
0.0015
-0.0462
Construct: Drought
Percent of county drought - extreme (+)
0.0066
0.0179
0.0008
0.0142
0.0084
Construct: Chemical Contamination
Arsenic (+)
0.1669
0.1674
0.1605
0.1584
0.1641
Barium (+)
0.1673
0.1684
0.1609
0.1628
0.1655
Cadmium (+)
0.1460
0.1475
0.1533
0.1615
0.1523
Chromium (+)
0.1661
0.1658
0.1592
0.1596
0.1636
Cyanide (+)
0.1369
0.1383
0.1181
0.1230
0.1291
Fluoride (+)
0.1736
0.1770
0.1804
0.1729
0.1765
Mercury (inorganic) (+)
0.0634
0.0494
0.0478
0.0614
0.0575
Nitrate (+)
0.1666
0.1600
0.1485
0.1417
0.1565
Nitrite (+)
0.1356
0.1322
0.1212
0.1231
0.1298
Selenium (+)
0.1661
0.1740
0.1644
0.1626
0.1663
Antimony (+)
0.1639
0.1541
0.1538
0.1586
0.1597
Endrin (+)
0.1392
0.1369
0.1387
0.1480
0.1412
Methoxychlor (+)
0.1670
0.1650
0.1676
0.1752
0.1690
Dalapon (+)
0.1462
0.1444
0.1409
0.1473
0.1449
Di(2-ethylhexyl) adipate (+)
0.1614
0.1576
0.1568
0.1624
0.1605
Simazine (+)
0.1674
0.1635
0.1651
0.1666
0.1671
Di(2-ethylhexyl) phthalate (+)
0.1682
0.1607
0.1594
0.1580
0.1638
Picloram (+)
0.1344
0.1301
0.1308
0.1445
0.1350
Dinoseb (+)
0.1599
0.1570
0.1550
0.1591
0.1584
Atrazine (+)
0.1758
0.1747
0.1738
0.1763
0.1759
2,4-Dichlorophenoxyacetic acid (+)
0.1612
0.1695
0.1565
0.1671
0.1623
Benzo[a]pyrene (+)
0.1578
0.1510
0.1538
0.1589
0.1561
Pentrachlorophenol (+)
0.1652
0.1622
0.1689
0.1715
0.1674
Polychlorinated biphenyls (+)
0.1244
0.1169
0.1081
0.1189
0.1185
1,2,-Dibromo-3-chloropropane (+)
0.1606
0.1552
0.1622
0.1631
0.1613
Ethylene dibromide (+)
0.0947
0.1043
0.1051
0.1035
0.1000
Xylenes (+)
0.1685
0.1654
0.1790
0.1816
0.1744
Chlordane (+)
0.1734
0.1755
0.1755
0.1763
0.1751
Dichloromethane (+)
0.1877
0.1950
0.1986
0.1900
0.1921
p-Dichlorobenzene (+)
0.1814
0.1886
0.1807
0.1814
0.1820
1,1,1-Trichloroethane (+)
0.1885
0.1917
0.1977
0.1906
0.1920
Trichloroethylene (+)
0.1893
0.1954
0.1992
0.1914
0.1932
Carbon tetrachloride (+)
0.1919
0.1968
0.2008
0.1926
0.1951
Benzene(+)
0.1880
0.1957
0.2008
0.1901
0.1929
33
-------
Table 9. continued.
Water Domain
Metropolitan-
Urbanized
(RUCC1 = 1167)
Nonmetropolitan
Urbanized (RUCC2 =
306)
Less Urbanized
(RUCC3 = 1026)
Thinly Populated
(RUCC4 = 644)
Total (All =
3143)
Toluene (+)
0.1839
0.1736
0.1908
0.1876
0.1859
Styrene (+)
0.1822
0.1927
0.1980
0.1905
0.1896
Alpha particles (+)
0.0670
0.0537
0.0609
0.0771
0.0639
cis1,2-Dichloroethylene (+)
0.1892
0.1958
0.1998
0.1904
0.1930
Total coliform proportion (+) 0.0084 -0.0088 0.0008 0.0105 0.0067
Land Domain
The loadings for the variables that comprise the mines construct
of the land domain varied by RUCC (Table 11), but loadings
for the variables that comprise the other constructs (agriculture,
pesticides, radon, and facilities) were consistent across RUCCs.
Each variable again has been annotated with a "+" or that is
the predicted direction for the loading to ensure that higher values
of the EQI represent worse enviromnental quality.
Table 11. Variable loadings, valence determination of variables - Land domain
Land Domain
Metropolitan-
Urbanized
(RUCC1 = 1167)
Nonmetropolitan
Urbanized
(RUCC2 = 306)
Less Urbanized
(RUCC3 = 1026)
Thinly Populated
(RUCC4 = 644)
Total
(All = 3143)
Construct: Agriculture
Farms per acre (+)
0.3742
0.3148
0.3275
0.3501
0.3487
Irrigated acreage (+)
0.2750
0.1364
0.1789
0.1720
0.2109
Chemicals used to control nematodes (+)
0.3127
0.2753
0.2883
0.3297
0.3070
Manure (+)
0.3701
0.3049
0.3174
0.3561
0.3483
Chemicals used to control diseases in crops and
orchards (+)
0.3589
0.3384
0.3302
0.3420
0.3479
Chemicals used to defoliate/control growth/thin fruit
M
0.2796
0.2486
0.2630
0.3209
0.2793
Harvested acreage (+)
0.4173
0.3943
0.4039
0.4074
0.4156
Animal units (+)
0.1876
0.1135
0.1118
0.1603
0.1479
Construct: Pesticides
Fungicides (+)
0.1055
0.2088
0.2125
0.0972
0.1582
Herbicides (+)
0.2007
0.3285
0.3177
0.2388
0.2742
Insecticides (+)
0.1759
0.2893
0.2604
0.1676
0.2272
Construct: Mines
Primarily coal mines, mines per county population (+)
-0.0220
-0.0497
-0.0966
-0.0583
-0.0611
Primarily metal mines, mines per county population (+)
-0.0836
-0.2283
-0.1961
-0.2172
-0.1754
Primarily nonmetal mines, mines per county
population (+)
0.0076
-0.0798
-0.0904
-0.0676
-0.0521
Primarily sand and gravel mines, mines per county
population (+)
0.1181
-0.0229
-0.0341
0.0058
0.0270
Primarily stone mines, mines per county population (+)
0.0740
-0.0971
-0.1101
-0.1088
-0.0515
Construct: Radon
Radon zone (+)
-0.0680
-0.0838
-0.0517
-0.1475
-0.0827
Construct: Facilities
Facilities (+)
0.1389
0.2361
0.1930
0.1322
0.1598
34
-------
Sociodemographic Domain
The loadings for the variables that comprise the
sociodemographic domain varied by RUCC (Table 12), indicating
some variables were more influential on the domain score in
urban counties, whereas others exerted more of an effect in rural
counties. For instance, percent unemployed loaded on the RUCC
1 sociodemographic domain at 0.16 compared with its loading
on RUCC 4 sociodemographic domain of 0.44. Each variable has
been annotated with a "+" or that is the predicted direction
for the loading. Because we want to ensure that higher values
of the EQI are associated with worse enviromnental quality,
those variables that we anticipate being associated with poor
enviromnental quality are assigned a "+" indicating more of this
attribute would be a negative for health. Most of the variables
initially loaded in nearly the opposite direction intended. The
loadings are a function of the program's starting point, or seed,
which is not easily manipulable. Therefore, the loading valence
needed to be corrected prior to the construction of the indices to
ensure that higher values on a given index, and on the overall
EQI, signify worse enviromnental quality. One important item to
note is that the patterns of association within the socioeconomic
construct across RUCC levels were not consistent. For instance,
percent Democratic voting in the 2008 election loaded negatively
in the most urban counties (RUCC 1 and 2) but positively in
the less urban counties (RUCC 3 and 4). Percent of individuals
earning a bachelor's degree, percent unemployed, percent of
families in poverty, median household value, and creative class
are variables that loaded in a consistent direction across rural-
urban strata. Appendix V provides the original and modified
valence corrected variable loadings.
Table 12. Valence corrected variable loadings, valence determination of variables - Sociodemographic domain
Sociodemographic Domain
Socioeconomic Construct
Metropolitan-
Urbanized
(RUCC1 = 1167)
-0.4689
0.1625
0.2591
0.2306
-0.4034
-0.3700
0.0055
-0.1827
-0.1162
Nonmetropolitan-
Urbanized
(RUCC2 = 306)
-0.4621
0.3274
0.4293
-0.1331
-0.4002
-0.0874
0.1371
0.0141
0.1604
Less
Urbanized
Thinly
Populated
OVERALL
(RUCC3 = 1026) (RUCC4 = 644) (n=3143)
-0.4174
0.3546
0.4737
-0.0555
-0.3476
-0.0640
0.1116
0.1523
0.2725
-0.4416
0.4418
0.4904
-0.1381
-0.2216
0.2578
-0.0141
0.0603
0.2766
-0.4585
0.1269
0.298
0.1979
-0.4331
-0.3824
0.1085
-0.1458
0.0118
-0.0094
0.2386
0.2997
0.2012
-0.0234
-0.4668
-0.4463
-0.3829
-0.2458
-0.4833
-0.2625
-0.0929
0.0374
0.2313
-0.211
Percent bachelors degree (-)
Percent unemployed (+)
Percent families less than poverty level (+)
Percent vacant housing (+)
Median household value (-)
Household income (-)
Count of occupants per room (+)
Percent renter-occupied housing (+)
Gini coefficient (+)
Crime Construct
Log violent crime (+)
Creative class construct
Creative class (-)
2008 Political valence construe
Percent Democratic (-)
35
-------
Built Domain
Similar to the sociodemographic domain, the loadings for the
variables that comprise the built domain varied by RUCC (Table
13), indicating some variables were more influential on the
domain score in urban counties, whereas others exerted more
of an effect in rural counties. Each variable again has been
annotated with a "+" or that is the predicted direction for the
loading to ensure that higher values of the EQI represent worse
enviromnental quality. Also, similar to the sociodemographic
domain, many of the initial variable loadings are opposite to that
intended. These loading valences needed to be valence corrected
prior to the construction of the indices to ensure that higher
values on a given index, and on the overall EQI, signify worse
enviromnental quality. The business-related enviromnents loaded
consistently across RUCC levels, as did the public transportation,
commute time and walkability score (Table 13). Appendix V
provides the original and modified valence corrected variable
loadings.
Table 13. Valence corrected variable loadings, valence determination of variables - Built domain
Metropolitan- Nonmetropolitan- Less Thinly
Urbanized Urbanized Urbanized Populated
Built Domain (RUCC1 = 1167) (RUCC2 = 306) (RUCC3 = 1026) (RUCC4 = 644) OVERALL (n=3143)
Socioeconomic Construct
Vice-related environment (+) -0.2676 -0.0331 -0.2724 -0.2595 -0.2930
Civic-related environment (-) -0.1238 -0.2057 -0.1890 -0.3102 -0.3071
Education-related environment (-) -0.2409 -0.2626 -0.3278 -0.3285 -0.3495
Health care-related environment (-) -0.4189 -0.3856 -0.3179 -0.2742 -0.2798
Negative food environment (+) -0.3239 -0.2707 -0.2306 -0.1527 -0.2280
Positive food environment (-) -0.3405 -0.2752 -0.2660 -0.2524 -0.3179
Recreation environment (-) -0.2354 -0.3484 -0.3212 -0.3222 -0.3590
Social service-related environment (-) -0.3446 -0.3503 -0.3644 -0.2793 -0.3629
Highway safety construct
Traffic fatality rate (+) -0.1978 0.2340 0.2197 0.2312 0.1751
Housing construct
Rate of low-rent + Section 8 housing (+) 0.1230 -0.0459 -0.0697 0.0178 -0.0581
Road construct
Proportion of secondary roads (+) -0.0950 0.1319 0.1761 0.2054 0.1777
Commuting behavior construct
Commute time (+) 0.1886 0.2808 0.3230 0.3546 0.3329
Public transportation (-) -0.2253 -0.1111 -0.0777 -0.0256 -0.0463
Walkability construct
Walkability score (-) -0.3516 -0.3310 -0.3542 -0.3787 -0.1585
Green space construct
Proportion green space (-) 0.1065 -0.0253 0.0418 0.1370 0.0451
Changes to 2006-2010 index construction from
original 2000-2005 EQI
Valence Assignment
The sole modification to the PCA methodology in the county
2006-2010 EQI compared to that of the 2000-2005 EQI is
"valence correction." We also have created a 2000-2005 valence
corrected version of the EQI.
The loading pattern for the air domain which is comprised of
established pollutants, served as the reference for our index
orientation. The vast majority of variables for the air domain
loaded "+" for both the overall United States and across the
rural-urban continuum. Thus, orientation for valence correction,
if needed, was toward variables with known poor enviromnental
attributes toward "+" loadings. Valence correction was applied
only to the sociodemographic and built-enviromnent domains.
This is because only the sociodemographic and built domains
had variables that were assigned as poor enviromnental attributes
that loaded initially as For instance, we were reasonably
certain that a high percentage of unemployed per county (variable
in sociodemographic domain) is anticipated to have deleterious
effects (and, therefore, could be assigned a "+" loading sign
based on our determined index orientation). Appendix V provides
the modified loadings, when applicable, along with the rationale
for valence correction.
Comparison of 2000-2005 EQI to the 2000-2005 valence
corrected EQI
To assess the impact of valence correction, we computed Pearson
and Spearman correlation coefficients between the nonvalence-
corrected and valence-corrected 2000-2005 EQI. For the overall
EQI, both the Pearson and Spearman correlation coefficients
were roughly 1. For RUCC1. they were 0.99 across both. For
RUCC2, the Pearson correlation coefficient was 0.99, whereas
36
-------
the Spearman correlation coefficient was 0.98. For RUCC3, the
Pearson and Spearman correlation coefficients were -0.97 and
-0.96, respectively. And, finally, forRUCC4, they were -0.97 and
-0.97, respectively.
Comparison of 2000-2005 valence corrected EQI to the
2006-2010 EQI
We additionally computed Pearson and Spearman correlation
coefficients between the valence corrected 2000-2005 EQI and
the 2006-2010 EQI. The domain-specific loadings for the overall
EQI differed over the two time periods in terms of magnitude,
rank, and direction. These differential loadings contributed to
the relatively low correlation between the 2000-2005 and 2006-
2010 periods. For the overall EQI, the Pearson and Spearman
correlation coefficients were both 0.34. For RUCC1, they were
-0.71 and -0.72, respectively. ForRUCC2, the Pearson correlation
coefficient was -0.35, whereas the Spearman correlation
coefficient was -0.37. For RUCC3, the Pearson and Spearman
correlation coefficients were 0.64 and 0.69, respectively. And,
finally, forRUCC4, they were 0.57 and 0.59, respectively. The
loadings may have differed over the two time periods because
of inputs that were included in the domains, valence correction
procedures, and potential changes in enviromnental quality. It
is for these reasons that we recommend the two indices not be
compared over time.
Domain-Specific Index Description and Loadings on
Overall EQI
The means, standard deviations, and ranges for each domain-
specific index are presented in Table 14. As expected, the index
loadings on the overall EQI index were mean (0) and standard
deviation (1). In examining the ranges of each RUCC-stratified
index, the larger the negative number (the smaller the minimum),
the better the enviromnental quality, whereas the larger the
maximum value, the worse the enviromnental quality. In general,
higher values of each domain's index was found in the more
metropolitan areas, and the maximum values went down as
counties became more thinly populated.
Table 14. Description of the domain indices contributing to the overall and rural-urban continuum codes (RUCCs) stratified
Environmental Quality Index for 3143 U.S. counties (2006-2010)
-4.39E-10
-6.72
3.71
Air Environment Index
-9.70E-10
-4.54
Land Environment Index
-2.11E-11
-5.13
2.76
Sociodemographic Environment Index
-2.20E-10
-7.29
Air Environment Index
1.28E-09
-4.30
Land Environment Index
-7.23E-10
-4.28
2.78
Sociodemographic Environment Index
-2.96E-09
-2.92
2.37
Air Environment Index
-2.11E-09
1.62
Land Environment Index
-1.45E-10
-4.14
Sociodemographic Environment Index
I.32E-10
-2.67
3.31
Air Environment Index
7.79E-10
Land Environment Index
7.34E-10
-4.79
3.64
Sociodemographic Environment Index
1.40E-09
-5.69
2.17
Air Environment Index
5.36E-10
-4.32
Land Environment Index
-1.17E-09
-3.51
3.81
Sociodemographic Environment Index
Built-Environment Index
-2.34E-09
-3.50
3.28
Built-Environment Index
-4.06E-10
-2.64
4.20
Water Environment Index
1.30E-10
-1.21
1.96
Less Urbanized (n=1026)
Metropolitan-Urbanized (n=1167)
Water Environment Index
-1.38E-09
1.93
Water Environment Index
-3.48E-12
-1.46
2.05
Built-Environment Index
6.18E-10
-3.22
3.77
Water Environment Index
2.94E-10
-3.95
2.37
Built-Environment Index
1.20E-09
-4.71
5.66
Built-Environment Index
-1.93E-09
-3.62
7.29
Water Environment Index
-1.59E-09
-1.61
1.56
Thinly Populated (n=644)
Non-Metropolitan-Urbanized (n=306)
All Counties (n=3143)
Mean
Standard Deviation
Minimum
Maximum
Mean Standard Deviation Minimum Maximum
All Counties (n=3143)
Air Environment Index -4.39E-10 1 -6.72 3.71
Water Environment Index -3.48E-12 1 -1.46 2.05
Land Environment Index -9.70E-10 1 -4.54 1.84
Built-Environment Index 1.20E-09 1 -4.71 5.66
Sociodemographic Environment Index -2.11E-11 1 -5.13 2.76
Metropolitan-Urbanized (n=1167)
Air Environment Index -2.20E-10 1 -7.29 3.68
Water Environment Index -1.38E-09 1 -1.48 1.93
Land Environment Index 1.28E-09 1 -4.30 1.80
Built-Environment Index -1.93E-09 1 -3.62 7.29
Sociodemographic Environment Index -7.23E-10 1 -4.28 2.78
Non-Metropolitan-Urbanized (n=306)
Air Environment Index -2.96E-09 1 -2.92 2.37
Water Environment Index -1.59E-09 1 -1.61 1.56
Land Environment Index -2.11E-09 1 -3.86 1.62
Built-Environment Index -2.34E-09 1 -3.50 3.28
Sociodemographic Environment Index -1.45E-10 1 -4.14 2.84
Less Urbanized (n=1026)
Air Environment Index 8.32E-10 1 -2.67 3.31
Water Environment Index 2.94E-10 1 -3.95 2.37
Land Environment Index 7.79E-10 1 -3.88 1.61
Built-Environment Index 6.18E-10 1 -3.22 3.77
Sociodemographic Environment Index 7.34E-10 1 -4.79 3.64
Thinly Populated (n=644)
Air Environment Index 1.40E-09 1 -5.69 2.17
Water Environment Index 1.30E-10 1 -1.21 1.96
Land Environment Index 5.36E-10 1 -4.32 1.51
Built-Environment Index -4.06E-10 1 -2.64 4.20
Sociodemographic Environment Index -1.17E-09 1 -3.51 3.81
37
-------
Description of Overall EQI
The pattern of association for the domain-specific loadings
differed by rural-urban status (Table 15). In the most urban areas,
RUCC1, the sociodemographic and built-enviromnent domains
were both influential, as indicated by their loading values (0.68
and 0.67, respectively), followed by the land domain (0.23).
For the nomnetropolitan-urbanized areas (RUCC2), the built
and sociodemographic domains loaded similarly on the overall
EQI (0.58 and 0.53, respectively), followed more closely by
the air domain. In all but the overall EQI, the water domain
was least influential, based on its low PCA coefficients. In the
most thinly populated counties, RUCC4, the water and land
domains were characterized by the lowest loadings (0.13 and
0.14, respectively), whereas the built, sociodemographic, and air
domains were the most influential (loadings of 0.60, 0.56, and
0.54, respectively).
The built and the air domains loaded approximately equally on
the overall EQI, and, unlike the loadings observed on the RUCC-
stratified EQIs, the sociodemographic domain was relatively
unimportant to the overall quality. Similar to the loadings for
each domain, the loadings for each RUCC-stratified EQI was
valence corrected to ensure that a higher EQI score corresponds
to worse enviromnental quality. Appendix VI contains county
mapping of the overall EQI 2006-2010 and RUCC-stratified
domain-specific indices.
Table 15. Loadings of the domain indices contributing to the overall and rural-urban continuum codes (RUCCs) stratified Environmental
Quality Index for 3143 U.S. counties (2006-2010)
Overall (n=3143)
Air Domain
Water Domain
Land Domain
Built-Environment Domain
Sociodemographic Domain
Metropolitan-Urbanized RUCC1 (n=1167)
Air Domain
Water Domain
Land Domain
Built-Environment Domain
Sociodemographic Domain
Nonmetropolitan Urbanized Areas RUCC 2 (n=306)
Air Domain
Water Domain
Land Domain
Built-Environment Domain
Sociodemographic Domain
Less Urbanized Areas RUCC 3 (n=1026)
Air Domain
Water Domain
Land Domain
Built-Environment Domain
Sociodemographic Domain
Thinly Populated RUCC 4 (n=644)
Air Domain
Water Domain
Land Domain
Built-Environment Domain
Sociodemographic Domain
Coefficient/Loading
0.6678
0.2209
0.3038
0.6240
-0.1536
-0.1280
-0.0906
0.2340
0.6730
0.6839
0.4128
-0.2407
0.3926
0.5274
0.5825
0.4785
-0.1569
0.1769
0.6370
0.5562
0.5402
0.1323
0.1430
0.5960
0.5612
95% Confidence Interval
0.6238,0.7118
0.0940, 0.3479
0.2054, 0.4021
0.5582, 0.6898
-0.2966,-0.0107
-0.2414, -0.0146
-0.2522,0.7010
0.0856, 0.3824
0.6377, 0.7083
0.6476, 0.7201
0.2771,0.5484
-0.4204, -0.0611
0.2514, 0.5337
0.4136, 0.6414
0.4939, 0.6712
0.4049, 0.5520
-0.2693, -0.0445
0.0672, 0.2866
0.5939, 0.6802
0.4939,0.6184
0.4809, 0.5994
0.0177,0.2469
0.0233, 0.2627
0.5469, 0.6450
0.5064,0.6160
38
-------
4.0
Discussion
This report describes the efforts to update the Environmental
Quality Index (EQI) for all counties in the United States for
the 2006-2010 period. The EQI was created for two main
purposes: (1) as an indicator of ambient conditions/exposure in
environmental health modeling and (2) as a covariate to adjust
for ambient conditions in environmental models. However, with
the public release of the EQI and variables that constructed the
EQI, other uses may emerge. The methods applied provide a
reproducible approach that capitalizes almost exclusively on
publicly available data sources.
The EQI holds promise for improving the environmental
estimation in public health. The EQI describes the ambient
county-level conditions to which residents are exposed,
whether they are at home, at school, or at work, provided
these multiple human activity spaces occur in the same county.
Since the creation of the EQI 2000-2005, multiple studies have
been conducted examining the relationship between overall
environmental quality and health outcomes, including preterm
birth[3], mortality[4], cancer incidence[5], asthma prevalence[6],
physical inactivity and obesity [7], infant mortality [8], and
pediatric multiple sclerosis[9], A complete list of references
related to EQI and health outcomes is listed in Appendix I.
With the updated EQI 2006-2010, the hope is that the EQI can
continue to be used to help public health researchers investigate
cumulative impact of various diverse constructs that typically
are viewed in isolation. Each of the domain-specific pieces of
information that contributes to the EQI is also informative.
Because most environmental health practice occurs on a domain-
specific basis, this domain-specific information may be important
to policymakers and environmental health practitioners. The
domain-specific loadings to the EQI indicate which of the
environmental domains accounts for the largest portion of the
variability in the EQI; in essence, these loadings answer the
question about which domain is making the biggest contribution
to the total environment. In addition, the variable loadings on
each of the domains are also informative for the same reason.
The development of the EQI 2006-2010 followed mostly the
same protocol as the EQI 2000-2005. Most of the constructs and
the data sources identified for each of the five domains in the EQI
2000-2005 were maintained. Principal components analysis was
used to develop the indices. However, using lessons learned from
the creation of the EQI 2000-2005, some modifications were
adopted to improve the EQI 2006-2010.
Summary of changes made to 2006-2010 version
compared with 2000-2005
Modifications to the EQI 2006-2010 included exploring new
data sources that were not available during EQI 2000-2005
development, assessment of all variables for continued inclusion
in the EQI, and assessment of variables' valence within a domain
and valence correction. Although most constructs were carried
over from the EQI 2000-2005 to the updated EQI 2006-2010,
the exceptions to this were the following: One deletion each in
the water domain and land domain and constructs added to the
water domain, land domain, sociodemographic domain, and the
built-environment domain. For data sources, we added seven new
data sources and discontinued use of one data source. Lastly, we
assessed the valence of each domain to ensure that the orientation
of the PCA output would have uniformity for interpretation of the
domain indices and uniformity for orientation as input into the
second PCA.
Strengths and Limitations
Because modifications were made to the updated EQI 2006-2010,
direct comparisons between EQI 2000-2005 and EQI 2006-2010
should not be made. The two indices should not be examined
as being continuous over time (e.g., if a study period covers
2004-2007, only one of the indices should be chosen or study
population should be stratified by time period matched to the
appropriate EQI).
The EQI offers a comprehensive measure of environmental
quality for all counties in the United States and is comprised of
many of the best environmental measures currently available.
The EQI can be used as an ambient exposure metric to help
identify environmental issues related to community health. It
provides information on overall environmental exposures faced
in a community. In addition, because data sources were used for
all U.S. counties, the EQI is comparable across counties to help
identify areas of better and worse overall environmental quality.
The development of domain-specific indices enables counties
to assess the drivers of poor environmental quality in their
county. Additionally, because it is comparable across counties,
areas that are burdened most by poor environmental quality
can be identified. Finally, the EQI can be used in a variety of
environmental health research activities as a control variable to
adjust for overall environmental exposure, while trying to isolate
a specific effect. Such a control variable will provide better
estimates of effects by reducing confounding by co-occurring
environmental factors.
The EQI is a national-level index that potentially can provide
a better understanding into how multiple environmental
conditions affect U.S. counties. At its current county-level
scale, the EQI may not reveal environmental injustices seen
39
-------
at the local community level. However, it does highlight those
counties experiencing an increased burden of environmental
impacts. Further, the EQI can contribute to environmental justice
endeavors by describing the process by which EQI data were
obtained and how the EQI was constructed and by indicating the
Web sites containing available data that can be used to construct
indices at different levels of aggregation.
The EQI can be a tool for interested investigators to consider
constructing local EQIs and adding relevant local-level data for
more focused comparisons.
Use of the EQI as a measure of exposure assumes exposure to
"environment" is consistent for all individuals, but the extent of
individual environmental exposure was not assessable. The EQI
was focused solely on the outside environment, which may not
be the most relevant exposure in relation to human health and
disease. Finally, population-level analyses offer little predictive
utility for individual-level risk. Therefore, although the index
may be useful at identifying less healthy county environments,
it will not be useful for predicting individual-level adverse
outcomes.
The EQI was developed for research purposes and is not meant
to be a diagnostic tool. The EQI would be useful to identify
potential areas of concern for counties to target future research,
but it should not be used to target regulatory purposes.
Data
Data sources evaluated represented each of the five
environmental domains. Each data source was reasonably well
documented. Despite finding a considerable number of data
sources applicable to each environmental domain, significant data
gaps exist.
The data used to create the index balanced quality measurement
with geographic breadth of coverage. Therefore, the index
does a solid job estimating the ambient environment but may
be less useful for estimating specific environments (e.g., in a
particular noncounty location in the United States at a specific
time). Not all relevant environmental exposures necessarily
were included in the index. Data inclusion was dependent on
data collection and coverage; if relevant data were not being
collected, the information was not captured in the EQI. Relatedly,
in areas where little data collection occurs, the data may be
overrepresenting the environmental profile of those areas. For
example, a county that contains a National Park without data
collected and a town with data collection will be represented
solely by the town data, although that may be inaccurate for
the entire county. Conversely, environments with a wealth of
environmental measurements, like urban areas, will be better
estimated by the EQI.
Environmental data sources often are plagued by inadequate
spatial and temporal coverage. Most of the data sources obtained
for the EQI required spatial interpolation to achieve county-
level estimates. For example, even with extensive air monitoring
networks, the measured spatial coverage of the United States was
incomplete, particularly in rural areas. Some types of measures
were located disproportionately in urban areas (e.g., PM air
pollution), whereas other sorts are found in rural areas (e.g.,
industrial livestock operations). The nonrandom distribution
of environmental risk meant that virtually all interpolated data
were inaccurate, impairing the assessment of how pollutants
differentially impacted urban and rural areas.
From a human health perspective, probably the biggest limitation
to existing environmental data sources is that data are collected
with little thought given to potential health impacts. For instance,
monitoring sites may collect relevant air pollutant data, but
their location (e.g., air monitors located on top of buildings)
is inappropriate for assessing the street-level values to which
humans are exposed. Pesticide data, from the land domain,
usually reports pesticide sales in relation to crops and livestock,
not application, handling, or disbursement. Even the United
States Census, which is widely used in health research, primarily
is collected for tax and political districting purposes. Some of
the data sources identified have not been used in human health
research and, as such, are a limitation. Regularly collected, high-
quality data that considers probable human health impacts would
make the task of assessing differential exposures considerably
easier.
Environmental data also were collected rarely with adequate
temporal frequency. Although data on some parameters were
collected on a consistent and frequent basis, the majority were
not. Water data, for instance, were collected only sporadically
in response to a particular query or based on regulatory statute.
Within the sociodemographic domain, the complete United States
Census was collected decennially, which limits investigators'
capacity to explore temporal changes. Some characteristics of
places can change rapidly, but, under current data collection
schedules, these changes cannot be assessed. Initially, the EQI
sought to estimate yearly measures. However, ultimately, only
the 5-year (2006-2010) and 6-year (2000-2005) measures
were created because of the lack of yearly data for some of the
variables.
Many environmental parameters were compiled at a smaller
unit of aggregation (e.g., for a municipality or city), and most
were not maintained in a single source, such as a data repository.
Although national repositories for some domains exist (e.g.,
water, air), often in response to federal regulations, no built-
environment repository exists (for transit, walkability/physical
activity, street connectivity, presence of sidewalks, or pedestrian
lighting measures). Localities with limited funds may not be
motivated or able to collect these data.
PCA Methodology
The use of PCA was not without limitations. Normality is
an important assumption for PCA, and not all the data were
distributed normally in their raw form. Many of the nonnormal
variables were those with a substantial number of meaningful
zeros (e.g., there were no public housing units contained
within these counties). This "absence" of attribute is important
information to convey, and, yet, it was problematic from a
score-construction perspective. Although transforming the
data improved their distribution, it reduced each variable's
interpretability. A PCA-derived score also can be challenging to
interpret. Outliers in the data also can be a limitation. However,
with 3143 counties and normality checks, this is less problematic
in the EQI.
40
-------
Although limited, the use of PCA was also an important strength
of this project. PCA provided a means to overcome one of
the significant limitations in the field of environmental health
and combine multiple environmental domains into one index
of ambient environmental quality; the whole endeavor would
not have been possible without this data reduction strategy.
The resulting scale is standardized, which will facilitate its
comparison to other scales constructed in different countries or at
different units of aggregation. Further, it is the approach that has
been used in other scale or score construction activities[65, 66],
Conclusion
The updated EQI2006-2010 was constructed for all counties
(n=3143) in the United States, incorporating data for five
environmental domains, (1) air, (2) water, (3) land, (4) built,
and (5) sociodemographic, and stratified by RUCCs. Mostly, the
same reproducible approach used to create EQI 2000-2005 also
was used to create EQI 2006-2010, with some noted changes
that incorporate lessons learned from the first version. The EQI
will be used as a measure in environmental health research. This
broad-based effort acknowledges the many factors that together
impact environmental quality and, more generally, recognizes
that these factors work together to impact public health. Updates
to the EQI for future years are planned, and the research team is
actively creating a census tract version as a first step to explore
other, finer spatial aggregations.
41
-------
-------
5.0
References
1. United States Environmental Protection Agency (EPA),
Creating an Overall Environmental Quality Index - Technical
Report. 2014. National Health and Environmental Effects
Research Laboratory: Chapel Hill, NC.
2. United States Environmental Protection Agency (EPA), EPA's
2008 Report on the Environment. 2008: Washington, DC.
3. Rappazzo, K.M., et al., The associations between
environmental quality and preterm birth in the United States,
2000-2005: A cross-sectional analysis. Environ Health, 2015.
14: p. 50.
4. Jian, Y., et al., Associations between Environmental Quality
and Mortality in the Contiguous United States, 2000-2005.
Environ Health Perspect, 2017. 125(3): p. 355-362.
5. Jagai, J.S., et al., County-level cumulative environmental
quality associated with cancer incidence. Cancer, 2017.
123(15): p. 2901-2908.
6. Gray, C.L., et al., Associations between environmental quality
and adult asthma prevalence in medical claims data. Environ
Res, 2018. 166: p. 529-536.
7. Gray, C.L., et al., The association between physical inactivity
and obesity is modified by five domains of environmental
quality in U.S. adults: A cross-sectional study. PLoS One,
2018. 13(8): p. e0203301.
8. Patel, A.P, et al., Associations between environmental quality
and infant mortality in the United States, 2000-2005. Arch
Public Health, 2018. 76: p. 60.
9. Lavery, A.M., et al., Examining the contributions of
environmental quality to pediatric multiple sclerosis. Mult
Scler Relat Disord, 2017.18: p. 164-169.
10. United States Environmental Protection Agency (EPA). Air
Quality System Data Mart. The Ambient Air Monitoring
Program. 2010.
11. United States Environmental Protection Agency (EPA),
National Air Toxics Assessments. 2005.
12. United States Environmental Protection Agency (EPA),
Watershed Assessment, Tracking, and Environmental Results
(WATERS). 2010.
13. Program, N.A.D., National Atmospheric Deposition
Program. 2010.
14. United States Geological Survey (USGS), Estimated Use of
Water in the United States. 2010.
15. United States Drought Monitor (USDM), Drought Monitor
Data Downloads. 2010.
16. United States Environmental Protection Agency (EPA),
National Contaminant Occurrence Database (NCOD). 2005.
17. United States Environmental Protection Agency (EPA), Safe
Drinking Water Information System. 2010.
18. Stone, W.W., Estimated annual agricultural pesticide use for
counties of the conterminous United States, 1992-2009. 2013,
U.S. Geological Survey.
19. United States Department of Agriculture (USD A), 2007
Census of Agriculture full report. 2009.
20. United States Environmental Protection Agency (EPA), EPA
Geo spatial Data Download Service. 2017.
21. United States Environmental Protection Agency (EPA), Map
of radon zones. 2017.
22. United States Department of Labor Mines Safety Health
Administration (MSHA), Mines Data Set. 2017.
23. United States Geological Survey (USGS), National
Geochemical Survey. 2006.
24. Bureau, U.S.C., American FactFinder. 2017.
25. Federal Bureau of Investigation (FBI), Uniform Crime
Reports. 2014.
26. Leip, D., Dave Leip's Atlas of U.S. Presidential Elections.
2016.
27. United States Department of Agriculture (USD A), Economic
Research Service (ERS) Creative Class County Codes. 2017.
28. Bradstreet, D.a., Dun andBradstreet Products. 2017.
29. Bureau, U.S.C., Topologically Integrated Geographic
Encoding and Referencing. 2017.
30. HERE. NAVTEQ traffic mapping. 2019 [cited 2019 April 2];
Available from: https://www, here. co m/iiavtea.
31. National Highway Traffic Safety Administration (NHTSA),
N.C.f.S.a.A.N., Fatality Analysis Reporting System (EARS).
2017.
32. Development, U.S.D.o.H.a.U., MultifamilyAssistance and
Section 8 Contracts Database. 2017.
33. United States Environmental Protection Agency (EPA),
EnviroAtlas Green space dataset. 2017.
34. Homer, C., et al., Completion of the 2011 National Land
Cover Database for the Conterminous United States-
representing a decade of land cover change information.
2015. 81(5): p. 345-354.
35. National Oceanic and Atmospheric Administration, O.f.C.M.,
Coastal Change Analysis Program (C-CAP) Regional Land
Cover. 2017.
36. United States Environmental Protection Agency (EPA),
National Walkability Index (NWI). 2017.
43
-------
37. United States Environmental Protection Agency (EPA).
National Emissions Inventory. 2019 [cited 2019 April
2]; Available from: https://www.epa.gov/air-emissions-
iiiventories/tiational-eiriissiotis-inveiitoiy-nei.
38. United States Geologic Services (USGS). National
Hydrography Dataset. 2019 [cited 2019 April 2]; Available
from: https://www.iisgs.gov/core-science-svstems/iigp/
natio nal-lwdro graphv.
39. United States Environmental Protection Agency (EPA).
Reach Address Database. 2010 [cited 2013 May 31];
Available from: http://www.epa.gov/waters/doc/rad/index.
html.
40. United States Environmental Protection Agency (EPA).
EPA Report on the Environment. 2019 [cited 2019 April 2];
Available from: https ://www. epa. gov/repo rt-env iro nment.
41. Mult-Resolution Land Cover Characteristics (MRLC)
Consortium. 2019 [cited 2019 April 2]; Available from:
https ://www. mrlc. gov/.
42. Cressie, N., The origins ofkriging. Mathematical Geology,
1990. 22(3): p. 239-252.
43. Tabachnick, B.G., Fidell, L.S., Using Multivariate Statistics.
5th ed. 2007, Boston: Pearson Allyn and Bacon.
44. Clean Water Act of1972.
45. United States Environmental Protection Agency (EPA).
National Pollutant Discharge Elimination System (NPDES).
December 12, 2018 [cited 2019 April 3]; Available from:
46. Kellog, R.L., Lander, C.H., Moffitt, D.C., Gollehon. N.,
Manure Nutrients Relative to the Capacity of Cropland and
Pasture land to Assimilate Nutrients: Spatial and Temporal
Trends for the United States. 2000, United States Department
of Agriculture.
47. Baker, N.T., Stone, W.W., Estimated annual agricultural
pesticide use for counties of the conterminous United
States, 2008-12. 2014: U.S. Department of the Interior, U.S.
Geological Survey.
48. United States Environmental Protection Agency (EPA).
Assessment, Cleanup, and Redevelopment Exchange
(ACRES) Brownfield Sites. 2010 [cited 2010 August 26];
Available from: http://www.epa.gov/browmieMs/.
49. United States Environmental Protection Agency (EPA).
Superfund National Priorities List (NPL) Sites. 2010;
Available from: http://www.epa.gov/siiperfiiiid/sites/iipl/
index, htm.
50. United States Environmental Protection Agency (EPA).
Section Seven Tracking System (SSTS) Pesticide Producing
Site Locations. 2019 [cited 2019 April 3]; Available from:
51. United States Environmental Protection Agency (EPA).
Resource Conservation and Recovery Act (RCRA) Large
Quantity Generators (LQG). 2010 [cited 2010 August 26];
Available from: http://www.epa.gov/osw/liazard/generation/
52. United States Environmental Protection Agency (EPA).
Resource Conservation and Recovery Act (RCRA) Treatment,
Storage, and Disposal Facilities (TSD) and (RCRA)
Corrective Action Facilities. 2010 [cited 2010 August 26];
Available from: http://www.epa.gov/osw/hazard/tsd/index.
htm.
53. National Technical Information Service. Federal Information
Processing Standards Publications (FIPS PUBS), [cited 2013
August 1]; Available from: http://www.mst.gov/itl/fips.cfm.
54. Richardson, E.A., et al., Green cities and health: a question
of scale? J Epidemiol Community Health, 2012. 66(2): p.
160-5.
55. Access, G.B.D.H., et al., Healthcare Access and Quality
Index based on mortality from causes amenable to personal
health care in 195 countries and territories, 1990-2015:
A novel analysis from the Global Burden of Disease Study
2015. Lancet, 2017. 390(10091): p. 231-266.
56. Friesen, C.E., Seliske, P., Papadopoulos, A., Using principal
component analysis to identify priority neighbourhoods for
health services delivery by ranking socioeconomic status.
2016. 8(2).
57. \fyas, S., Kumaranayake, L., Constructing socio-economic
status indices: how to use principal components analysis.
2006. 21(6): p. 459-468.
58. Jolliife, I.T., Cadima, J., Principal component analysis: a
review and recent developments. Philos Trans A Math Phys
Eng Sci, 2016. 374(2065): p. 20150202.
59. Hall, S.A., Kaufman, J.S., Ricketts, T.C., Defining urban
and rural areas in U.S. epidemiologic studies. J Urban
Health, 2006. 83(2): p. 162-75.
60. United States Department of Agriculture (USD A). Measuring
rurality: Rural-urban continuum codes, [cited 2019 April
3]; Available from: https://www.eis.iisda.gov/data-products/
riiral-iirbari-cofititiiiiiin-codes//.
61. Langlois, PH., et al., Occurrence of conotruncal heart birth
defects in Texas: a comparison of urban/rural classifications.
J Rural Health, 2010. 26(2): p. 164-74.
62. Langlois, PH., et al., Urban versus rural residence and
occurrence of septal heart defects in Texas. Birth Defects Res
A Clin Mol Teratol, 2009. 85(9): p. 764-72.
63. Luben, T.J., et al., Urban-rural residence and the occurrence
of neural tube defects in Texas, 1999-2003. Health Place,
2009. 15(3): p. 848-54.
64. Messer, L.C., et al., Urban-rural residence and the
occurrence of cleft lip and cleft palate in Texas, 1999-2003.
Ann Epidemiol, 2010. 20(1): p. 32-9.
65. Emerson, J., et al., 2012 Environmental Performance Index
and Pilot Trend Environmental Performance Index - Full
Report. 2012, Yale Center for Environmental Law and Policy:
New Haven, CT.
66. Messer, L.C., et al., The development of a standardized
neighborhood deprivation index. J Urban Health, 2006.
83(6): p. 1041-62.
44
-------
Appendix I: List of References Related
to 2000-2005 Environmental Quality Index
1. Lobdell DT, Jagai JS, Rappazzo K, Messer LC. (2011)
Data sources for environmental assessment: determining
availability, quality and utility, American Journal of Public
Health Suppl l:S277-85.
2. Jagai JS, Rosenbaum BJ, Pierson SM, Messer LC, Rappazzo
K, Naumova EN, Lobdell DT. (2013) Putting Regulatory
Data to Work at the Service of Public Health: Utilizing
Data Collected Under the Clean Water Act. Water Quality,
Exposure, and Health 5:117-125; DOI: 10.1007/sl2403-013-
0095-1.
3. Messer LC, Jagai JS, Rappazzo KM, Lobdell DT. (2014)
Construction of an environmental quality index for public
health research. Environmental Health 13:39; DOI:
10.1186/1476-069X-13-39.
4. Rappazzo KM, Messer LC, Jagai JS, Gray CL, Grabich SC,
Lobdell DT. (2015) The association between environmental
quality and preterm birth in the United States, 2000-2005: a
cross-sectional analysis. Environmental Health 14:50;DOI:
10.1186/sl2940-015-0038-3.
5. Grabich SC, Horney J, Konrad C, Lobdell DT. (2015).
Measuring the Storm: Methods of Quantifying Hurricane
Exposure with Pregnancy Outcomes. Natural Hazards
Review; DOI: 10.1061/(ASCE)NH. 1527-6996.0000204.
6. Grabich SC, Rappazzo KM, Gray CL, Jagai JS, Jian Y,
Messer LM, Lobdell DT. (2016) Additive interaction between
heterogeneous environmental quality domains (air, water,
land, sociodemographic, and built environment) on preterm
birth. Frontiers in Public Health, http://dx.doi.org/10.3389/
fpubh.2016.00232.
7. Jian Y, Messer LC, Jagai JS, Rappazzo KM, Gray CL,
Grabich SC, Lobdell DT. (2017) The associations between
environmental quality and mortality in the contiguous United
States 2000-2005. Environmental Health Perspectives
125:355-362, http://dx.doi.org/10.1289/EHP119.
8. Jagai JS, Messer LC, Rappazzo KM, Gray CL, Grabich SC,
Lobdell DT. (2017) County-level cumulative environmental
quality associated with cancer incidence. Cancer, http://
dx. do i. o r g/10.1002/cncr. 30709.
9. Lavery AM, Waldman AT, Charles Casper T, Roalstad S,
Candee M, Rose J, Belman A, Weinstock-Guttman B, Aaen
G, Tillema JM, Rodriguez M, Ness J, Harris Y, Graves J,
Krupp L, Benson L, Gorman M, Moodley M, Rensel M,
Goyal M, Mar S, Chitnis T, Schreiner T, Lotze T, Greenberg
B, Kahn I, Rubin J, Waubant E; U.S. Network of Pediatric
MS Centers. (2017) Examining the contributions of
environmental quality to pediatric multiple sclerosis. Multiple
Sclerosis and Related Disorders 18:164-169, https://doi.
org/10.1016/j .msard.2017.09.004.
10. Jian Y, Wu CYH, Go hike JM. (2017) Effect modification by
environmental quality on the association between heatwaves
and mortality in Alabama, United States. International
Journal of Environmental Research and Public Health
14:1143, https://doi.org/10.3390/iieiDhl4101143.
11. Gray CL, Lobdell DT, Rappazzo KM, Jian Y, Jagai JS,
Messer LC, Patel AP, DeFlorio-Barker SA, Lyttle C, Solway
J, Rzhetsky A. (2018) Associations between environmental
quality and adult asthma prevalence in medical claims
data. Environmental Research 166:529-536, https://doi.
org/10.1016/j .envres.2018.06.020.
12. Gray CL, Messer LC, Rappazzo KM, Jagai JS, Grabich
SC, Lobdell DT. (2018) The association between physical
inactivity and obesity is modified by five domains of
environmental quality in U.S. adults: A cross-sectional study.
PLoS One, https://doi.org/10.1371/iournal.Pone.02Q3301.
13. Patel AP, Jagai JS, Messer LC, Gray CL, Rappazzo KM,
Deflorio-Barker SA, Lobdell DT. (2018) Associations
between environmental quality and infant mortality in the
United States, 2000-2005. Archives of Public Health 76:60,
https://doi.org/10.1186/sl3690-018-0306-0.
14. Kosnik MB, Reif DM, Lobdell DT, Astell-Burt T, Feng X,
Hader JD, Hoppin JA. (2019) Associations between access to
healthcare, environmental quality, and end-stage renal disease
survival time: proportional-hazards models of over 1,000,000
people over 14 years. PLoS One, fattps://doi.org/10.1371/
journal.pone.0214094.
15. Jagai JS, Krajewski AK, Shaikh S, Lobdell DT, Sargis
RM. (2020) Association between environmental quality
and diabetes in the USA. Journal of Diabetes Investigation
11 (2):315-324, https://doi.org/10.llll/idi.13152.
16. Huanga M, Xiaob J, Nasca PC, Liu C, Lu Y, Lawrence WR,
Wang L, Chen Q, Lin S. (2019) Do multiple environmental
factors impact four cancers in women in the contiguous
United States? Environmental Research 179:108782, https://
doi.ofg/10.1016/i.envres.2019.108782.
17. Wang M, Wasserman E, Geyer N, Carroll RM, Zhao S, Zhang
L, Hohl R, Lengerich EJ, McDonald AC. (2020) Spatial
patterns in prostate cancer-specific mortality in Pennsylvania
using Pennsylvania Cancer registry data, 2004-2014.
A-l
-------
18. Gearhart-Serna LM, Hoffman K. Devi GR. (2020)
Environmental Quality and Invasive Breast Cancer.
Cancer Epidemiology, Biomarkers & Prevention; DOI:
10.1158/1055-9965.EPI-19-1497.
19. Li X, Xiao J, Huang M, Liu T, Guo L, Zeng W, Chen
Q, Zhang J, Ma W. (2020) Associations of county-level
cumulative environmental quality with mortality of chronic
obstructive pulmonary disease and mortality of tracheal,
bronchus, and lung cancers. Science of the Total Environment
703:135523, https://doi.Org/10.1016/i.scitoteiiv.2019.135523.
A-2
-------
Appendix II: Identified Variables by Source
for Each Domain
Variables by Data Source - Air Domain
AIR QUALITY SYSTEM (AQS)
Variable
Particulate Matter <10 micrometers in aerodynamic
diameter (PM10)
Particulate Matter <2.5 micrometers in aerodynamic
diameter (PM2.5)
Nitrogen Dioxide (N02)
Sulfur Dioxide (S02)
Ozone (03)
Carbon Monoxide (CO)
Variable Name
ln_S02
ln_NOx
ln_CO
PM25
PM10
03
Counties/Monitors
3143/1187
3143/1146
3143/303
3143/499
3143/575
3143/442
(jg/m3
ppm, log transformed
ppb, log transformed
PPb
ppm, log transformed
Variable Notes
|jg/m3
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
EQI Version
NATIONAL AIR TOXICS ASSESSMENT (NATA)
NOTES: WHEN DATA IS MISSING/NOT RECORDED, ZERO VALUES WERE DEEMED APPROPRIATE. MOST VARIABLES KEPT FOR EQI HAVE BEEN LOG TRANSFORMED.
EQI 2006-2010 = NATA 2005. ALL VARIABLES REPORTED IN TONS EMITTED PER YEAR. UNLESS OTHERWISE NOTED, ALL VARIABLES ARE LOG TRANSFORMED.
VARIABLES WERE DROPPED DUE TO INSUFFICIENT DATA (HIGH NUMBERS OF MISSING OR ZERO OBSERVATIONS) OR DUE TO HIGH CORRELATION WITH OTHER
VARIABLES.
Variable Variable Name Counties Variable Notes EQI Version
1,1,2,2-tetrachloroethane A_TeCA_ln 3137 2000-2005; 2006-2010
1,1,2-trichloroethane A_112TCA_ln 3137 2000-2005; 2006-2010
1.2-dibromo-3-chloropropane A_DBCP_ln 3137 2000-2005; 2006-2010
1.3-dichloropropene A_DCI_propene_ln 3061 2006-2010
Acrylic acid A_Acrylic_acid_ln 3107 2000-2005; 2006-2010
Benzidine A_Benzidine_ln 3137 2000-2005; 2006-2010
Benzyl chloride A_Benzyl_CI_ln 3137 2000-2005; 2006-2010
Beryllium compounds A_Be_ln 3137 2000-2005; 2006-2010
bis-2-ethylhexyl phthalate A_DEHP_ln 3137 2000-2005; 2006-2010
Carbon tetrachloride A_CCI4 3137 2000-2005; 2006-2010
Carbonyl sulfide A_CylS_ln 3137 2006-2010
Chlorine A_CI_ln 3137 2000-2005; 2006-2010
Chlorobenzene A_C6H5CI_ln 3137 2000-2005; 2006-2010
Chloroform A_chloroform_ln 3137 2000-2005; 2006-2010
Chloroprene A_Chloroprene_ln 3137 2000-2005; 2006-2010
Chromium compounds A_Cr_ln 3137 2000-2005; 2006-2010
Cobalt compounds A_Co_ln 3132 2006-2010
Cyanide compounds A_CN_ln 3137 2000-2005; 2006-2010
Dibutylphthalate A_DBP_ln 3137 2000-2005; 2006-2010
Ethyl chloride A_EtCI_ln 3136 2000-2005; 2006-2010
Ethylbenzene A_Ebenzine 3137 2006-2010
Ethylene dibromide A_EDB 3137 2000-2005; 2006-2010
Ethylene dichloride A_EDC_ln 3137 2000-2005; 2006-2010
Formaldehyde A_Formaldehyde 3137 2006-2010
Glycol ethers A_Glycol_ethers_ln 3057 2000-2005; 2006-2010
Hydrazine A_N2H2_ln 3137 2000-2005; 2006-2010
B-l
-------
Variable
Hydrochloric acid
Isophorone
Manganese compounds
Methyl bromide
Methylene chloride
Phosphine
Polychlorinated biphenyls
Propylene dichloride
Quinoline
Trichloroethylene
Vinyl chloride
Variable Name
A_HCI_ln
AJsophoroneJn
A_Mn_ln
A_Me_Br_ln
A_MeCI2_ln
A_PH3_ln
A_PCBs_ln
A_ProCI2_ln
A_Quinolin_ln
A_C2HCI3_ln
A_VyCI_ln
Counties
3137
3131
3137
3137
3137
3062
3137
3137
3137
3137
3137
Variable Notes
EQI Version
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
Variables by Data Source - Water Domain
WATERS PROGRAM DATABASE/REACH ADDRESS DATABASE
NOTES: THESE MEASURES WERE COMPUTED; LOTS OF MISSING DATA, SO SEVERAL VARIABLES CANNOT BE USED. VARIABLES CALCULATED USING REACH STREAM
LENGTH DATABASE. DATA FOR 2006, 2008, AND 2010 WERE AVERAGED. DATA WAS UPDATED BASED ON 2010 FIPS CODES.
Variable
Percent of stream length
impaired in county
All NPDES Permits grouped
per 1000km of stream
length in county
Variable Name
D303_Percent
ALLNPDESperKM
Counties
2513
3141
Variable Notes
Calculated with REACH
database information
I types of NPDES Permits
EQI Version
2000-2005;
2006-2010
2006-2010
Notes
Grouped variable of Sewage Permits per 1000 km of
Stream in County; Industrial Permits per 1000 km of
Stream in County; Stormwater Permits per 1000 km of
Stream in County
ESTIMATE USE OF WATER IN THE UNITED STATES
NOTES: THESE MEASURES WERE COMPUTED FOR 2005 AND 2010 DATA AND AVERAGED. USGS PROVIDES ESTIMATES AT COUNTY LEVEL, SO NO ADDITIONAL
MANIPULATION REQUIRED.
Variable
Percent ofPopulation on SelfSupply, 2005, 2010
Percent of Public Supply Population that is on Surface Water, 2005, 2010
Variable Name
Per_TotPopSS
Per_PSWithSW
Counties Variable Notes
3141 Estimate provided at county level
3067 Estimate provided at county level
NATIONAL ATMOSPHERIC DEPOSITION PROGRAM
NOTES: MEASURES PROVIDED AT VARIOUS MONITORING STATIONS. VALUES FOR 2006-2010 WERE KRIGED TO NATIONAL LEVEL COVERAGE. DATA FOR ALL YEARS
WAS AVERAGED TOGETHER.
Variable
Variable Name
Counties
Variable Notes
EQI Version
Calcium (Ca) precipitation weighted mean (mg/L)
CaAveJn
3141
Kriged & log transformed
2000-2005; 2006-2010
Potassium (K) precipitation weighted mean (mg/L)
KAveJn
3141
Kriged & log transformed
2000-2005; 2006-2010
Nitrate (N03) precipitation weighted mean (mg/L)
N03Ave
3141
Kriged - transformation not needed
2000-2005; 2006-2010
Chloride (CI) deposition
ClAveJn
3141
Kriged & log transformed
2000-2005; 2006-2010
Sulfate (S04) deposition
S04Ave_ln
3141
Kriged & log transformed
2000-2005; 2006-2010
Total Mercury deposition (ng/M2)
Use only values with A or B quality rating
HgAve
3141
Kriged - transformation not needed
2000-2005; 2006-2010
DROUGHT MONITOR DATA
NOTES: RASTER DATA AGGREGATED TO THE COUNTY LEVEL. DATA FOR ALL YEARS 2006-2010 WAS AVERAGED TOGETHER.
Variable Variable Name Counties
Percent of county drought-extreme (D3-D4) AvgOfD3_ave 3141
Variable Notes
EQI Version
2000-2005; 2006-2010
B-2
-------
NATIONAL CONTAMINANT OCCURRENCE DATABASE (NCOD)
NOTES: WILL USE 6 YEAR REVIEW 2 (DATA COLLECTED BETWEEN 1998-2005).
CALCULATE THE FOLLOWING VARIABLES FOR EACH CHEMICAL FOR EACH COUNTY (AGGREGATING ALL PWS IN COUNTY) FOR ALL YEARS COMBINED; MISSING FOR
THOSE COUNTIES WITHOUT ANY DATA; DID NOT KEEP DETECTS.
1990
2000-2005; 2006-2010
Barium - average
W_Ba_ln (mg/L)
Average for all samples in county, log transformed
1989
2000-2005; 2006-2010
Chromium (total) - average
W_Cr_ln (mg/L)
Average for all samples in county, log transformed
2138
2000-2005; 2006-2010
Fluoride - average
W_FL_ln (mg/L)
Average for all samples in county, log transformed
W_N03_ln (mg/L)
2000-2005; 2006-2010
Nitrate (as N) - average
Average for all samples in county, log transformed
1986
2000-2005; 2006-2010
Selenium - average
W_SE_ln (mg/L)
Average for all samples in county, log transformed
1509
2000-2005; 2006-2010
Endrin - average
W_Endrin_ln (ug/L)
Average for all samples in county, log transformed
1292
2000-2005; 2006-2010
Dalapon - average
W_Dalapon_ln (ug/L)
Average for all samples in county, log transformed
1669
2000-2005; 2006-2010
Simazine - average
Di(2-ethylhexyl) phthalate (DEHP)
W_Simazine_ln (ug/L)
Average for all samples in county, log transformed
1430
2000-2005; 2006-2010
Benzo[a]pyrene - average
W_BenzoAP_ln (ug/L)
Average for all samples in county, log transformed
Polychlorinated biphenyls (PCBs)
- average
2000-2005; 2006-2010
W_PCB_ln (ug/L)
Average for all samples in county, log transformed
Ethylene dibromide (EDB) -
average
1630
2000-2005; 2006-2010
W_EDB_ln (ug/L)
Average for all samples in county, log transformed
1498
2000-2005; 2006-2010
Chlordane - average
W_Chlordane_ln (ug/L)
Average for all samples in county, log transformed
1,4-Dichlorobenzene
(p-Dichlorobenzene) - average
2165
2000-2005; 2006-2010
W_PDCB_ln (ug/L)
Average for all samples in county, log transformed
2250
2000-2005; 2006-2010
Trichloroethylene - average
W_Trichlorenejn (ug/L)
Average for all samples in county, log transformed
Monochlorobenzene
Cyanide - average
W_CN_ln (mg/L)
1385
Average for all samples in county, log transformed
2000-2005; 2006-2010
Mercury (inorganic) - average
W_HG_ln (mg/L)
2056
Average for all samples in county, log transformed
2000-2005; 2006-2010
Xylenes (Total) - average
W_xylenes_ln (ug/L)
2203
Average for all samples in county, log transformed
2000-2005; 2006-2010
Cadmium - average
W_Cd_ln (mg/L)
1991
Average for all samples in county, log transformed
2000-2005; 2006-2010
Nitrite (as N) - average
W_N02_ln (mg/L)
1583
Average for all samples in county, log transformed
2000-2005; 2006-2010
Antimony - average
W_Sb_ln (mg/L)
1994
Average for all samples in county, log transformed
2000-2005; 2006-2010
Methoxychlor - average
W_methoxychlor_ln (ug/L)
1512
Average for all samples in county, log transformed
2000-2005; 2006-2010
Pentachlorophenol - average
W_PCP_ln (ug/L)
1547
Average for all samples in county, log transformed
2000-2005; 2006-2010
1,1,1-Trichloroethane - average W_111 trichlorane_ln (ug/L)
2238
Average for all samples in county, log transformed
2000-2005; 2006-2010
Tetrachloroethylene - average
W_C2CI4_ln (ug/L)
224
Average for all samples in county, log transformed
2000-2005; 2006-2010
Di(2-ethylhexyl)adipate (DEHA)
- average
W_DEHA_ln (ug/L)
1456
Average for all samples in county, log transformed
2000-2005; 2006-2010
1,2-Dibromo-3-chloropropane
(DBCP) - average
W_DBCP_ln (ug/L)
1652
Average for all samples in county, log transformed
2000-2005; 2006-2010
2,4-D (2,4-Dichlorophenoxyacetic
acid) - average
W_24D_ln (ug/L)
1360
Average for all samples in county, log transformed
2000-2005; 2006-2010
Dichloromethane (Methylene
chloride) - average
W_DCM_ln (ug/L)
2245
Average for all samples in county, log transformed
2000-2005; 2006-2010
Alpha Particles (Gross Alpha,
excl.Radon&U) - average
W_alpha (PCI/L)
1243
Average for all samples in county
Variable
Arsenic - average
Variable Name
W_As_ln (mg/L)
Counties
2017
Variable Notes
Average for all samples in county, log transformed
EQI Version
2000-2005; 2006-2010
B-3
-------
SAFE DRINKING WATER INFORMATION SYSTEM (SDWIS)
NOTES: CUMULATIVE COUNT OF VIOLATIONS FOR ALL PWS IN COUNTY FOR THE YEAR. DATA IS AVAILABLE ANNUALLY DATA WERE COMPILED FOR 2006-2010.
Variable Variable Name Counties Variable Notes EQI Version
Total Coliform, Proportion Coliform_Sum 2034 2006-2010
Variables by Source - Land Domain
2007 CENSUS OF AGRICULTURE
NOTES: ACRES OF CROP OR TREATMENT WERE DIVIDED BY TOTAL COUNTY ACRES TO GET PERCENTAGE OF ITEM PER COUNTY. SOME COUNTIES HAD
SUPPRESSED ACREAGE DUE TO IDENTIFIABILITY ISSUES. FOR THESE, THE UNACCOUNTED-FOR ACREAGE FOR EACH STATE WAS CALCULATED (TOTAL STATE
ACREAGE - LISTED COUNTY ACREAGE). THE ACREAGE WAS DIVIDED EQUALLY AMONG THE FARMS IN COUNTIES WITH SUPPRESSED INFORMATION. DATA FOR
HAWAII AND ALASKA ARE NOT AVAILABLE. THESE DATAARE REFRESHED EVERY 5 YEARS. THE NEXT AVAILABLE DATA IS FOR 2012.
Variable Variable Name Counties Variable Notes EQI Version
Commercial fertilizer, lime, and soil conditioners pct_lime_acres 3065 2000-2005; 2006-2010
Manure
Chemicals used to control insects
Chemicals used to control weeds, grass, or
brush
Chemicals used to control nematodes
Chemicals used to control diseases in crops
and orchards
pct_manure_acres_ln 2975
pct_insecticide_acres 3141
pct_weed_acres 3061
pct_nematode_acres_ln 1933
pct_disease_acres_ln 2530
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2588
2000-2005; 2006-2010
Corn for grain (bushels)
pct_corn_acres
2082
2000-2005; 2006-2010
Soybeans for beans (bushels)
pct_soybean_acres
Potatoes (cwt)
Pct_potato_acres
1565
2000-2005; 2006-2010
Wheat for grain, all (bushels)
pct_wheat_acres
2520
2000-2005; 2006-2010
Chemicals used to control growth, thin fruit, or
defoliate
pct_defoliate_acres_ln
1980
2000-2005; 2006-2010
Animal units
Number of farms
Irrigated acres
Harvested acres
pd_au_ln
farms_per_acre_ln
pct_irrigated_acres_ln
pct_harvest_acres
3078
3039
2815
2755
1 AU is equal to 0.94 cattle and calves, 5.88 hogs and
pigs, 250 egg laying chickens, and 455 broiler chickens.
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2009 NATIONAL PESTICIDE USE DATASET (NPUD)
NOTES: PESTICIDE CONCENTRATIONS WERE GROUPED BY CLASS AND ADDED TOGETHER TO GET CLASS-LEVEL ESTIMATES OF PESTICIDE APPLICATION. THESE
DATAARE REFRESHED EVERY 5 YEARS. THE NEXT AVAILABLE DATA IS FOR 2012.
Variable Variable Name Counties Variable Notes EQI Version
Insecticides insecticidesjn 2761 2000-2005; 2006-2010
Herbicides herbicidesjn 2907 2000-2005; 2006-2010
Fungicides fungicidesjn 2256 2000-2005; 2006-2010
MAP OF RADON ZONE (EPA)
NOTES: THE EPA RADON ZONE MAP IDENTIFIES AREAS OF THE UNITED STATES WITH THE POTENTIAL FOR ELEVATED INDOOR RADON LEVELS. EACH UNITED
STATES COUNTY (3142) IS ASSIGNED TO ONE OF THREE ZONES BASED ON RADON POTENTIAL. DATA YEARS UNAVAILABLE. PRESUMABLY, RADON IS A STABLE
FEATURE, AND THE MAP IS NOT VARIABLE, BUT REFRESH DATES ARE NOT AVAILABLE. NO OTHER INFORMATION AVAILABLE IN DATA DOCUMENTATION.
Variable Variable Name Counties Variable Notes EQI Version
Radon zones Radon_zone 3142 3-level variable 2000-2005; 2006-2010
B-4
-------
SUPERFUND NATIONAL PRIORITIES LIST (NPL) SITES
NOTES: NPL SITE LOCATIONS AVAILABLE THROUGH THE EPAGEOSPATIAL DATA ACCESS PROJECT SITES WERE INCLUDED IN THE COUNTS IF THEY WERE
IDENTIFIED BETWEEN 2006-2010. PUBLISHED AUGUST 2016. START AND END DATES NOT AVAILABLE. DATA REFRESHED MONTHLY.
Variable Variable Name Counties Variable Notes EQI Version Notes
Included as part of composite
Count of Superfund National Priority List sites per county sf_county_count 719 2000-2005; 2006-2010 count variable
RESOURCE CONSERVATION AND RECOVERY ACT (RCRA) TREATMENT, STORAGE, AND DISPOSAL FACILITIES (TSD) AND RCRA
CORRECTIVE ACTION FACILITIES
NOTES: RCA TSD AND CORRECTION ACTION FACILITIES SITE LOCATIONS AVAILABLE THROUGH THE EPAGEOSPATIAL DATA ACCESS PROJECT. SITES WERE
INCLUDED IN THE COUNTS IF THEY WERE IDENTIFIED BETWEEN 2006-2010. PUBLISHED AUGUST 2016. START AND END DATES NOT AVAILABLE. DATA REFRESHED
MONTHLY.
Variable Variable Name Counties Variable Notes EQI Version Notes
Count of RCRA TSD and corrective Included as part of composite
action facilities per county rcra_tsd_count_by_fips 874 2000-2005; 2006-2010 count variable
RESOURCE CONSERVATION AND RECOVERY ACT (RCRA) LARGE QUANTITY GENERATORS (LQG)
NOTES: RCA LQG SITE LOCATIONS THROUGH THE EPA GEOSPATIAL DATA ACCESS PROJECT. SITES WERE INCLUDED IN THE COUNTS IF THEY WERE IDENTIFIED
BETWEEN 2006-2010. PUBLISHED AUGUST 2016. START AND END DATES NOT AVAILABLE. DATA REFRESHED MONTHLY.
Variable Variable Name Counties Variable Notes EQI Version Notes
Count of RCRA LQG facilities per county rcralqg_count 1963 2000-2005; 2006-2010 Included as part of composite count variable
TOXIC RELEASE INVENTORY (TRI) SITES
NOTES: TRI SITES AVAILABLE THROUGH THE EPA GEOSPATIAL DATA ACCESS PROJECT. SITES WERE INCLUDED IN THE COUNTS IF THEY WERE IDENTIFIED BETWEEN
2006-2010. PUBLISHED AUGUST 2016. START AND END DATES NOT AVAILABLE. DATA REFRESHED MONTHLY.
Variable Variable Name Counties Variable Notes EQI Version Notes
Count of TRI sites per county tri_county_count 2671 2000-2005; 2006-2010 Included as part of composite count variable
ASSESSMENT, CLEANUP, AND REDEVELOPMENT EXCHANGE (ACRES) BROWNFIELD SITES
NOTES: BROWNFIELD SITE LOCATIONS AVAILABLE THROUGH THE EPAGEOSPATIAL DATA ACCESS PROJECT. SITES WERE INCLUDED IN THE COUNTS IF THEY WERE
IDENTIFIED BETWEEN 2006-2010. PUBLISHED AUGUST 2016. START AND END DATES NOT AVAILABLE. DATA REFRESHED MONTHLY.
Variable Variable Name Counties Variable Notes EQI Version Notes
Count of ACRES sites per county acres_county_count 1273 2000-2005; 2006-2010 Included as part of composite count variable
SECTION SEVEN TRACKING SYSTEM (SSTS) PESTICIDE PRODUCING SITE LOCATIONS
NOTES: SSTS PESTICIDE-PRODUCING SITE LOCATIONS AVAILABLE THROUGH THE EPAGEOSPATIAL DATA ACCESS PROJECT. SITES WERE INCLUDED IN THE COUNTS
IF THEY WERE IDENTIFIED BETWEEN 2006-2010. PUBLISHED AUGUST 2016. START AND END DATES NOT AVAILABLE. DATA REFRESHED BUT NOT ANNUALLY.
Variable Variable Name Counties Variable Notes EQI Version Notes
Count of SSTS sites per county ssts_county_count 2099 2000-2005; 2006-2010 Included as part of composite count variable
MINE SAFETY AND HEALTH ADMINISTRATION (MSHA)
NOTES: THE MINE DATASET LISTS ALL COAL AND METAL/NON-METAL MINES UNDER MSHA'S JURISDICTION SINCE 1/1/1970. IT INCLUDES SUCH INFORMATION AS
THE CURRENT STATUS OF EACH MINE (ACTIVE, ABANDONED, NONPRODUCING, ETC.), THE CURRENT OWNER AND OPERATING COMPANY, COMMODITY CODES AND
PHYSICAL ATTRIBUTES OF THE MINE. MINE ID IS THE UNIQUE KEY FOR THIS DATA (https://ARLWEB.MSHA.GOV/OPENGOVERNMENTDATA/OGIMSHA.ASP). DATA
REFRESHED WEEKLY. COUNTIES WITH ZERO MINES WERE GIVEN A VALUE OF MINIMUM VALUE/2. THESE DATA WERE TRANSFORMED (LOG) TO ACCOUNT FOR THE
LARGE NUMBER OF ZEROS AND TO RESULT IN NEARLY NORMALLY DISTRIBUTED DATA.
Variable Variable Name Counties Variable Notes EQI Version Notes
Primarily coal mines, mines per county population Std_coal_prim_pop_ln 464 See notes above 2006-2010
Primarily metal mines, mines per county population Std_coal_prim_pop_ln 386 See notes above 2006-2010
Primarily nonmetal mines, mines per county population Std_coal_prim_pop_ln 1135 See notes above 2006-2010
Primarily sand and gravel mines, mines per county
population Std_coal_prim_pop_ln 2342 See notes above 2006-2010
Primarily stone mines, mines per county population Std_coal_prim_pop_ln 1965 See notes above 2006-2010
B-5
-------
Variables by Source - Sociodemographic Domain
UNITED STATES CENSUS SUMMARY FILES
NOTES: MANY, MANY MORE VARIABLES ARE AVAILABLE FROM THE UNITED STATES CENSUS THAN WILL BE DESCRIBED HERE. THE VARIABLES IDENTIFIED HERE
ARE THOSE THAT WILL BE USED IN THE EQI AND NOT THE PLETHORA OF VARIABLES THAT COULD BE CONSTRUCTED. DATA ARE AVAILABLE FOR MULTIPLE UNITS
OF GEOGRAPHIC AGGREGATION, INCLUDING THE COUNTY-LEVEL. FULL POPULATION DATA ARE COLLECTED DECENNIALLY; SAMPLE DATA ARE COLLECTED MORE
FREQUENTLY. DATA ARE AVAILABLE FOR DOWNLOAD FROM THE UNITED STATES CENSUS BUREAU WEB SITE.
Variable
Percent renter-occupied units
Percent vacant units
Median household value
Median household income
Bachelor's degree or higher,
percent of persons age 25 years+
Percent of persons who are
unemployed
Percent of families in poverty
Occupants per Room
Measure of income inequality
(proportion)
Variable Name
Pct_RenterOcc
Pct_Vacant_Housing
med_hh_value
ln_HH_lnc
Pct_BS
Pct_Unemp_total
Pct_Fam_Pov
ln_Occs_Room
GINLest
Counties
3143
3143
3143
3143
3143
3143
3143
3143
3143
Variable Notes EQI Version
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2000-2005; 2006-2010
2006-2010
2000-2005; 2006-2010
2006-2010
2006-2010
2006-2010
Notes
This variable replaced percent < HS
This variable replaced percent families in
poverty
This variable replaced number rooms / house
FBI UNIFORM CRIME REPORTS
NOTES: FBI UCR DATA WERE DOWNLOADED FOR EACH COUNTY IN EACH STATE FROM THE WEBSITE (HTTPS //WWW UCRDATATOOI GOV/) DATA ARE AVAILABLE
BY YEAR AND BY CRIME TYPE (VIOLENT = MURDER AND NONNEGLIGENT MANSLAUGHTER, FORCIBLE RAPE, ROBBERY, AND AGGRAVATED ASSAULT; PROPERTY
= BURGLARY, LARCENY-THEFT, AND MOTOR VEHICLE THEFT). DATA FROM 2006-2010 WERE TEMPORALLY AND SPATIALLY KRIGED FOR USE IN THE EQI. DATA
REPORTING IS VOLUNTARY. DATA ARE AVAILABLE AT THE CITY AND COUNTY LEVELS, BUT MANY COUNTIES DO NOT REPORT THESE DATA. DATA FOR LAW
ENFORCEMENT AGENCIES SERVING CITY JURISDICTIONS WITH POPULATIONS OF 10,000 OR MORE AND COUNTY AGENCIES OF 25,000 OR MORE. THEREFORE, DATA
MAY NOT BE AVAILABLE FOR EACH JURISDICTION EACH YEAR. DATA ARE AVAILABLE FROM 1960 TO CURRENT YEAR. RATES WERE OBTAINED FROM THE FBI. THE
VIOLENT CRIME RATE DATA WERE TRANSFORMED (LOG) TO ACCOUNT FOR THE LARGE NUMBER OF ZEROS AND TO RESULT IN NEARLY NORMALLY DISTRIBUTED
DATA.
Variable
Variable Name
Violent crime rate ln_ViolAv
Murder-manslaughter crime rate murder_manslaughter_rate
Rape crime rate rape_rate
Robbery crime rate rob_rate
Aggravated assault crime rate agg_assault_rate
Counties Variable Notes EQI Version Notes
Variable kriged to estimate values for counties 2000-2005;
3143 with no reported violent crime data 2006-2010
Variable kriged to estimate values for counties Constituent of violent
1062 with no reported violent crime data No crime rate
Variable kriged to estimate values for counties Constituent of violent
1055 with no reported violent crime data No crime rate
Variable kriged to estimate values for counties Constituent of violent
1062 with no reported violent crime data No crime rate
Variable kriged to estimate values for counties Constituent of violent
1062 with no reported violent crime data No crime rate
UNITED STATES DEPARTMENT OF AGRICULTURE ECONOMIC RESEARCH SERVICE CREATIVE CLASS INDEX
NOTES: THE ECONOMIC RESEARCH SERVICE (ERS) CLASS CODES INDICATE A COUNTY'S SHARE OF POPULATION EMPLOYED IN OCCUPATIONS THAT REQUIRE
"THINKING CREATIVELY." THIS SKILL ELEMENT IS DEFINED AS "DEVELOPING, DESIGNING, OR CREATING NEW APPLICATIONS, IDEAS, RELATIONSHIPS, SYSTEMS, OR
PRODUCTS, INCLUDING ARTISTIC CONTRIBUTIONS." DATA ARE AVAILABLE FOR DOWNLOAD FROM THE USDAERS WEBSITE.
Variable
Percent county employed in creative class
Variable Name
Num CreatClass
Counties
3143
Variable Notes
EQI Version
2006-2010
UNITED STATES ELECTION ATLAS
NOTES: THE POLITICAL CLIMATE OF A COUNTY WAS REPRESENTED BY THE DAVID LEIP ELECTION MAP. COUNTY-SPECIFIC PERCENTS VOTING REPUBLICAN OR
DEMOCRATIC WERE REPORTED. THE REPORT VOTING DEMOCRATIC IN THE 2008 PRESIDENTIAL ELECTION WERE INCLUDED IN THE EQI.
Variable
Percent county voting Democratic in 2008
Variable Name
DEM02008
Counties
3143
Variable Notes
EQI Version
2006-2010
B-6
-------
Variables by Source - Built Domain
HOUSING AND URBAN DEVELOPMENT (HUD) DATA
NOTES: THESE DATA PROVIDE A COUNT OF THE LOW-RENT AND SECTION 8 HOUSING IN EACH HOUSING AUTHORITY AREA. THESE HOUSING AUTHORITY AREAS
CORRESPOND TO CITIES, WHICH ARE THEN ASSIGNED FIPS CODES. COUNTIES WITHOUT HOUSING AUTHORITY CITIES ARE GIVEN A COUNT OF ZERO FOR LOW-
RENT AND/OR SECTION-EIGHT HOUSING. THESE DATA WERE TRANSFORMED (LOG) TO ACCOUNT FOR THE LARGE NUMBER OF ZEROS AND TO RESULT IN NEARLY
NORMALLY DISTRIBUTED DATA. DATA ARE REFRESHED FREQUENTLY), BUT UPDATE FREQUENCY NOT PROVIDED. HISTORIC DATA DOES NOT APPEAR TO BE
AVAILABLE FROM WEB SITE. DATA WERE COLLECTED IN 2010, BUT, SINCE LOW-RENT AND SECTION 8 HOUSING DOES NOT CHANGE SUBSTANTIALLY OVER TIME,
THESE DATA ARE CONSIDERED REPRESENTATIVE OF THE 2006-2010 TIME PERIOD. RATES FOR EACH VARIABLE CONSTRUCTED BY DIVIDING COUNT BY COUNTY
POPULATION.
Variable
Rate of low-rent + section 8
units in county
Count of low-rent units per
county
Count of section 8 units per
county
Variable Name Counties Variable Notes
Variable transformed (log) to allow it to
total_units_ln 3143 approximate normal distribution
Variable transformed (log) to allow it to
low_rent_units 2080 approximate normal distribution
Variable transformed (log) to allow it to
section_eight_units 2080 approximate normal distribution
EQI Version
2000-2005;
2006-2010
Notes
Zeros considered meaningful
zeros (lack of public housing)
Constituent of total unit rate
Constituent of total unit rate
FATALITY ANALYSIS REPORTING SYSTEM (FARS) DATA
NOTES: THE FATALITY ANALYSIS REPORTING SYSTEM (FARS) IS A NATIONWIDE CENSUS PROVIDING THE NATIONAL HIGHWAY TRAFFIC SAFETY ADMINISTRATION
YEARLY DATA REGARDING FATAL INJURIES SUFFERED IN MOTOR VEHICLE TRAFFIC CRASHES. FARS DATAARE AVAILABLE FROM 1975 (HTTPV/WWW.NHTSA.GOV/
FARS I). RATES FOR THE COUNT OF FATAL CRASHES PER COUNTY FOR 2006-2010 WERE CONSTRUCTED BY DIVIDING COUNT BY COUNTY POPULATION. THESE DATA
WERE TRANSFORMED (LOG) TO ACCOUNT FOR THE LARGE NUMBER OF ZEROS AND TO RESULT IN NEARLY NORMALLY DISTRIBUTED DATA. THESE DATA CAN BE
UPDATED ANNUALLY.
Variable
Variable Name Counties Variable Notes EQI Version Notes
Rate of fatal car crashes Variable transformed (log) to allow it to
per county ln_fatalities 3143 approximate normal distribution 2000-2005; 2006-2010
2010 UNITED STATES CENSUS SUMMARY FILES
NOTES: MANY, MANY MORE VARIABLES ARE AVAILABLE FROM THE UNITED STATES CENSUS THAN WILL BE DESCRIBED HERE. THE VARIABLES IDENTIFIED HERE
ARE THOSE THAT WILL BE USED IN THE EQI AND NOT THE PLETHORA OF VARIABLES THAT COULD BE CONSTRUCTED. DATAARE AVAILABLE FOR MULTIPLE UNITS
OF GEOGRAPHIC AGGREGATION, INCLUDING THE COUNTY-LEVEL. FULL POPULATION DATAARE COLLECTED DECENNIALLY; SAMPLE DATAARE COLLECTED MORE
FREQUENTLY. THESE DATA WERE TRANSFORMED (LOG) TO ACCOUNT FOR THE LARGE NUMBER OF ZEROS AND TO RESULT IN NEARLY NORMALLY DISTRIBUTED
DATA. DATAARE AVAILABLE FOR DOWNLOAD FROM THE UNITED STATES CENSUS BUREAU WEB SITE.
Variable Variable Name Counties Variable Notes EQI Version Notes
Percent of county residents who Variable transformed (log) to allow it to
report using public transportation ln_PubTrans 3143 approximate normal distribution 2000-2005; 2006-2010
Time it takes from home to go
to work CommuteTime 3143 Recorded in minutes 2006-2010
TIGER FILES
NOTES: TOPOLOGICALLY INTEGRATED GEOGRAPHIC ENCODING AND REFERENCING PRODUCTS PROVIDE MAPS AND ROAD LAYERS WORLDWIDE. INCLUDING THE
UNITED STATES. THESE DATAARE UPDATED REGULARLY BUT DO NOT CHANGE SUBSTANTIALLY OVER TIME. THE DATA USED IN THE EQI ARE FROM 2009. DATAARE
AVAILABLE AT CENSUS GEOGRAPHY. FOR THE STREET TYPES, THE HIGHWAY AND SECONDARY AND LOCAL ROADS (TERTIARY ROADS) PER COUNTY PER STATE
WERE DOWNLOADED. PROPORTION OF EACH ROAD TYPE WAS CONSTRUCTED BY DIVIDING THE DISTANCE OF EACH ROAD TYPE BY THE TOTAL AMOUNT OF EACH
ROAD.
Variable Variable Name Counties Variable Notes
Proportion of all roads that are
secondary roads SecondaryRoadProportion 3143
EQI Version Notes
This single variable replaced proportion
2006-2010 primary road and highways
B-7
-------
DUN AND BRADSTREET
NOTES: DUN AND BRADSTREET COLLECT COMMERCIAL INFORMATION ON BUSINESS. ITS DATABASE CONTAINS MORE THAN 195 MILLION RECORDS AND IS
PROPRIETARY. THE DATA ARE PUT THROUGH AN EXTENSIVE QUALITY ASSURANCE PROCESS, WHICH INCLUDES OVER 2000 SEPARATE AUTOMATED AND SEVERAL
MANUAL CHECKS. DATA ARE UPDATED DAILY. RATES OF EACH TYPE OF BUSINESS IN 2008 WERE CALCULATED BY DIVIDING THE COUNTS OF EACH VARIABLE BY
THE COUNTY POPULATION. THESE DATA WERE TRANSFORMED (LOG) TO ACCOUNT FOR THE LARGE NUMBER OF ZEROS AND TO RESULT IN NEARLY NORMALLY
DISTRIBUTED DATA.
Variable Variable Name Counties Variable Notes EQI Version Notes
Rate of positive food environment
businesses per county pos_food_rate_ln 3140 2000-2005; 2006-2010
Rate of negative food environment
businesses per county neg_food_rate_ln 3117 2000-2005; 2006-2010
Rate of alcohol, pawn, gaming
businesses per county al_pwn_gm_env_rate_ln 3039 2000-2005; 2006-2010
Rate of health care-related
businesses per county hc_env_rate_ln 3119 2000-2005; 2006-2010
Rate of recreation-related
businesses per county rec_env_rate_ln 3133 2000-2005; 2006-2010
Rate of education-related
businesses per county ed_env_rate_ln 3141 2000-2005; 2006-2010
Rate of social-service-related
businesses per county ss_env_rate_ln 3125 2000-2005; 2006-2010
Rate of civic-related businesses
per county civic_env_rate_ln 3138 2006-2010
ENVIROATLAS LAND COVER CONTERMINOUS UNITED STATES (EPA)
NOTES: THIS ENVIROATLAS DATASET REPRESENTS THE PERCENTAGE OF LAND AREA THAT IS CLASSIFIED AS NATURAL, BARREN, FOREST, TUNDRA, SHRUBLAND,
HERBACEOUS, WETLAND, WOODY WETLAND, EMERGENT WETLAND, ALL HUMAN LAND USE, DEVELOPED, OPEN SPACE DEVELOPED, LOW-INTENSITY DEVELOPED,
MEDIUM-INTENSITY DEVELOPED, HIGH-INTENSITY DEVELOPED, AGRICULTURAL, PASTURE/HAY, AND CULTIVATED CROP USING THE 2011 NATIONAL LAND COVER
DATASET (NLCD) FOR EACH COUNTY IN THE CONTERMINOUS UNITED STATES. THIS DATASET WAS PRODUCED BY THE UNITED STATES EPA TO SUPPORT
RESEARCH AND ONLINE MAPPING ACTIVITIES RELATED TO ENVIROATLAS. ENVIROATLAS (HTTPSV/WWW.EPA.GOV/ENVIROATLASI ENABLES THE USER TO INTERACT
WITH A WEB-BASED, EASY-TO-USE, MAPPING APPLICATION TO VIEW AND ANALYZE MULTIPLE ECOSYSTEM SERVICES FOR THE CONTIGUOUS UNITED STATES. THE
DATASET IS AVAILABLE AS DOWNLOADABLE DATA (HTTPSV/EDG.EPA.GOV/DATA/PUBLIC/ORD/ENVIROATLASl OR AS AN ENVIROATLAS MAP SERVICE. ADDITIONAL
DESCRIPTIVE INFORMATION ABOUT EACH ATTRIBUTE IN THIS DATASET CAN BE FOUND IN ITS ASSOCIATED ENVIROATLAS FACT SHEET (HTTPS://WWW.EPA.GOV/
ENVIROATLAS/ENVIROATLAS-FACT-SHEETS).
Variable Variable Name Counties Variable Notes EQI Version Notes
Combined natural land cover
and open space developed NINDEX_open 3109 Green space composite variable 2006-2010
Percentage of county land
area that is classified as
natural land cover
Percentage of county land
area that is classified as
barren land cover
Percentage of county land
area that is classified as
forest land cover
Percentage of county land
area that is classified as
tundra land cover
Percentage of county land
area that is classified as
shrubland land cover
Percentage of county land
area that is classified as
herbaceous land cover
Percentage of county land
area that is classified as
wetland land cover
Percentage of county land
area that is classified as
woody wetland land cover
NINDEX
pbar
pfor
ptun
pshb
phrb
pwtl
pwtlw
Composite variable of barren, forest,
tundra, shrubland, herbaceous, and wetland
3109 land cover 2006-2010
3109 Vegetation accounts for <15% total cover 2006-2010
Composite variable of deciduous,
evergreen, and mixed forests. Areas
dominated by trees generally greater than
5-meters tall, and greater than 20% total
3109 vegetation cover 2006-2010
3109 Alaska only areas 2006-2010
Areas dominated by shrubs; less than
5-meters tall; shrub canopy greater than
3109 20% of total vegetation 2006-2010
Areas dominated by graminoid and
herbaceous vegetation, usually greater than
3109 80% of total vegetation 2006-2010
Composite variable of woody and emergent
3109 wetlands. 2006-2010
Soil or substrate is periodically saturated
with or covered with water, and forest or
shrubland vegetation account for >20%
3109 vegetative cover 2006-2010
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
B-8
-------
Variable Variable Name Counties Variable Notes EQI Version Notes
Soil or substrate is periodically saturated
Percentage of county land with or covered with water, and perennial
area that is classified as herbaceous vegetation accounts for >80% Included as part of green space
emergent wetland land cover pwtle 3109 vegetative cover No composite variable
Percentage of county land
area that is classified as all Composite variable of developed and
human land use land cover UINDEX 3109 agricultural land cover No Does not meet definition of green space
Percentage of county land
area that is classified as
developed land cover pdev 3109 All developed land cover No Does not meet definition of green space
Percentage of county land
area that is classified as Mixture of some constructed materials
open space developed land but mostly vegetation; < 20% impervious Included as part of green space
cover pdevo 3109 surface No composite variable
Percentage of county land
area that is classified as
low-intensity developed land Mixture of constructed materials and
cover pdevl 3109 vegetation; 20% to 49% impervious surface No Does not meet definition of green space
Percentage of county land
area that is classified as
medium-intensity developed Mixture of constructed materials and
land cover pdevm 3109 vegetation; 50% to 79% impervious surface No Does not meet definition of green space
Percentage of county land
area that is classified as
high-intensity developed Highly developed areas; 80% to 100%
land cover pdevh 3109 impervious surface No Does not meet definition of green space
Percentage of county land
area that is classified as Composite variable of pasture/hay and
agricultural land cover pagr 3109 cultivated crop land cover No Does not meet definition of green space
Grasses, legumes, or grass-legume
Percentage of county land mixtures for livestock grazing; production of
area that is classified as seed or hay crops; pasture/hay vegetation
pasture/hay land cover pagrp 3109 accounts for >20% total vegetation No Does not meet definition of green space
Percentage of county land
area that is classified as
cultivated crop land cover
pagrc
3109
Production of annual crops; crop vegetation
accounts for >20% total vegetation;
includes land being actively tilled
No
Does not meet definition of green space
ENVIROATLAS LAND COVER ALASKA (EPA)
NOTES: THIS ENVIROATLAS DATASET REPRESENTS THE PERCENTAGE OF LAND AREA THAT IS CLASSIFIED AS NATURAL, BARREN, FOREST, TUNDRA, SHRUBLAND,
HERBACEOUS, WETLAND, WOODY WETLAND, EMERGENT WETLAND, ALL HUMAN LAND USE, DEVELOPED, OPEN SPACE DEVELOPED, LOW-INTENSITY DEVELOPED,
MEDIUM-INTENSITY DEVELOPED, HIGH-INTENSITY DEVELOPED, AGRICULTURAL, PASTURE/HAY, CULTIVATED CROP, AND PERENNIAL SNOW/ICE USING THE 2011
NATIONAL LAND COVER DATASET (NLCD) FOR EACH COUNTY IN ALASKA. THIS DATASET WAS PRODUCED BY THE UNITED STATES EPA TO SUPPORT RESEARCH AND
ONLINE MAPPING ACTIVITIES RELATED TO ENVIROATLAS. ENVIROATLAS (HTTPS7/WWW FPA GOV/FNVIROATI AS) ENABLES THE USER TO INTERACT WITH A WEB-
BASED, EASY-TO-USE, MAPPING APPLICATION TO VIEW AND ANALYZE MULTIPLE ECOSYSTEM SERVICES FOR THE CONTIGUOUS UNITED STATES. THE DATASET IS
AVAILABLE AS DOWNLOADABLE DATA (HTTPS7/FDG FPA GOV/DATA/PUBI IC/ORD/FNVIROATI AS) OR AS AN ENVIROATLAS MAP SERVICE. ADDITIONAL DESCRIPTIVE
INFORMATION ABOUT EACH ATTRIBUTE IN THIS DATASET CAN BE FOUND IN ITS ASSOCIATED ENVIROATLAS FACT SHEET (HHPS//WWW FPA GOV/FNVIROATI AS/
FNVIROATI AS-FACT-SHFFTS)
Variable
Combined natural land cover and
open space developed
Percentage of county land area that
is classified as natural land cover
Percentage of county land area that
is classified as barren land cover
Percentage of county land area that
is classified as forest land cover
Percentage of county land area that
is classified as tundra land cover
Variable Name Counties
NINDEX_open
NINDEX
pbar
pfor
ptun
29
29
29
29
29
Variable Notes EQIVersion
Green space composite variable 2006-2010
Composite variable of barren, forest, tundra,
shrubland, herbaceous, and wetland land
cover 2006-2010
Vegetation accounts for <15% total cover 2006-2010
Composite variable of deciduous, evergreen,
and mixed forests. Areas dominated by trees
generally greater than 5-meters tall, and
greater than 20% total vegetation cover 2006-2010
Alaska only areas; includes dwarf scrub,
sedge/herbaceous, lichens, and moss land
cover 2006-2010
Notes
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
B-9
-------
Variable Variable Name Counties Variable Notes EQIVersion Notes
Areas dominated by shrubs; less than
Percentage of county land area that 5-meters tall; shrub canopy greater than 20% Included as part of green space
is classified as shrubland land cover pshb 29 of total vegetation 2006-2010 composite variable
Percentage of county land area that Areas dominated by graminoid and
is classified as herbaceous land herbaceous vegetation, usually greater than Included as part of green space
cover phrb 29 80% of total vegetation 2006-2010 composite variable
Percentage of county land area that Composite variable ofwoody and emergent Included as part of green space
is classified as wetland land cover pwtl 29 wetlands 2006-2010 composite variable
Percentage of county land area Soil or substrate is periodically saturated with
that is classified as woody wetland or covered with water and forest or shrubland Included as part of green space
land cover pwtlw 29 vegetation account for >20% vegetative cover 2006-2010 composite variable
Soil or substrate is periodically saturated
Percentage of county land area that with or covered with water, and perennial
is classified as emergent wetland herbaceous vegetation accounts for >80% Included as part of green space
land cover pwtle 29 vegetative cover 2006-2010 composite variable
Percentage of county land area that
is classified as all human land use Composite variable of developed and Does not meet definition of green
land cover UINDEX 29 agricultural land cover No space
Percentage of county land area that Does not meet definition of green
is classified as developed land cover pdev 29 All developed land cover No space
Percentage of county land area
that is classified as open space Mixture of some constructed materials but Included as part of green space
developed land cover pdevo 29 mostly vegetation; <20% impervious surface No composite variable
Percentage of county land area
that is classified as low-intensity Mixture of constructed materials and Does not meet definition of green
developed land cover pdevl 29 vegetation; 20% to 49% impervious surface No space
Percentage of county land area that
is classified as medium-intensity Mixture of constructed materials and Does not meet definition of green
developed land cover pdevm 29 vegetation; 50% to 79% impervious surface No space
Percentage of county land area
that is classified as high-intensity Highly developed areas; 80% to 100% Does not meet definition of green
developed land cover pdevh 29 impervious surface No space
Percentage of county land area
that is classified as agricultural land Composite variable of pasture/hay and Does not meet definition of green
cover pagr 29 cultivated crop land cover No space
Grasses, legumes, or grass-legume mixtures
Percentage of county land area that for livestock grazing; production of seed or
is classified as pasture/hay land hay crops; pasture/hay vegetation accounts Does not meet definition of green
cover pagrp 29 for >20% total vegetation No space
Percentage of county land area that Production of annual crops; crop vegetation
is classified as cultivated crop land accounts for >20% total vegetation; includes Does not meet definition of green
cover pagrc 29 land being actively tilled No space
Percentage of county land area that
is classified as forest and woody Composite variable of forest and woody Included as part of green space
wetland cover Pfor90 29 wetland No composite variable
Percentage of county land area that
is classified as forest and emergent Included as part of green space
wetland cover Pwetl95 29 Composite of forest and emergent wetland No composite variable
Percentage of county land area that Characterized by perennial cover of ice and/ Does not meet definition of green
is classified as perennial snow/ice pice 29 or snow, generally >25% total cover No space
B-IO
-------
ENVIROATLAS LAND COVER HAWAII (EPA)
NOTES: THIS ENVIROATLAS DATASET REPRESENTS THE PERCENTAGE OF LAND AREA THAT IS CLASSIFIED AS NATURAL, BARREN, FOREST, TUNDRA, SHRUBLAND,
HERBACEOUS, WETLAND, WOODY WETLAND, EMERGENT WETLAND, ALL HUMAN LAND USE, DEVELOPED, OPEN SPACE DEVELOPED, LOW-INTENSITY DEVELOPED,
MEDIUM-INTENSITY DEVELOPED, HIGH -INTENSITY DEVELOPED, AGRICULTURAL, PASTURE/HAY, AND CULTIVATED CROP LAND COVER USING THE ENVIROATLAS
COMPOSITE OF THE 2005-2011 COASTAL CHANGE ANALYSIS PROGRAM (C-CAP) LAND COVER DATASET FOR EACH 12-DIGIT HYDROLOGIC UNIT CODE (HUC) IN
HAWAII. THIS DATASET WAS PRODUCED BY THE UNITED STATES EPA TO SUPPORT RESEARCH AND ONLINE MAPPING ACTIVITIES RELATED TO ENVIROATLAS.
ENVIROATLAS (HTTPSV/WWW.EPA.GOV/ENVIROATLAS) ENABLES THE USER TO INTERACT WITH A WEB-BASED, EASY-TO-USE, MAPPING APPLICATION TO VIEW AND
ANALYZE MULTIPLE ECOSYSTEM SERVICES FOR THE CONTIGUOUS UNITED STATES. THE DATASET IS AVAILABLE AS DOWNLOADABLE DATA (HTTPSV/EDG.EPA.GOV/
DATA/PUBLIC/ORD/ENVIROATLAS) OR AS AN ENVIROATLAS MAP SERVICE. ADDITIONAL DESCRIPTIVE INFORMATION ABOUT EACH ATTRIBUTE IN THIS DATASET CAN
BE FOUND IN ITS ASSOCIATED ENVIROATLAS FACT SHEET (HTTPSV/WWW.EPA.GOV/ENVIROATLAS/ENVIROATLAS-FACT-SHEETSV
Variable
Combined natural land cover and
open space developed
Percentage of county land area
that is classified as natural land
Percentage of county land area
that is classified as barren land
Percentage of county land area
that is classified as forest land
Percentage of county land area
that is classified as tundra land
cover
Percentage of county land area
that is classified as shrubland
land cover
Percentage of county land area
that is classified as herbaceous
land cover
Percentage of county land area
that is classified as wetland land
Percentage of county land
area that is classified as woody
wetland land cover
Percentage of county land area
that is classified as emergent
wetland land cover
Percentage of county land area
that is classified as all human
land use land cover
Percentage of county land area
that is classified as developed
land cover
Percentage of county land area
that is classified as open space
developed land cover
Percentage of county land area
that is classified as low-intensity
developed land cover
Percentage of county land area
that is classified as medium-
intensity developed land cover
Percentage of county land area
that is classified as high-intensity
developed land cover
Percentage of county land area
that is classified as agricultural
land cover
Variable Name Counties Variable Notes EQIVersion
NINDEX_open 5 Green space composite variable 2006-2010
Composite variable of barren, forest,
tundra, shrubland, herbaceous, and
NINDEX 5 wetland land cover 2006-2010
Vegetation accounts for <15% total
pbar 5 cover 2006-2010
Composite variable of deciduous,
evergreen, and mixed forests. Areas
dominated by trees generally greater
than 5-meters tall, and greater than 20%
pfor 5 total vegetation cover 2006-2010
ptun 5 Alaska only areas 2006-2010
Areas dominated by shrubs; less than
5-meters tall; shrub canopy greater than
pshb 5 20% of total vegetation 2006-2010
Areas dominated by graminoid and
herbaceous vegetation, usually greater
phrb 5 than 80% of total vegetation 2006-2010
Composite variable of woody and
pwtl 5 emergent wetlands 2006-2010
Soil or substrate is periodically saturated
with or covered with water and forest or
shrubland vegetation account for >20%
pwtlw 5 vegetative cover 2006-2010
Soil or substrate is periodically saturated
with or covered with water and perennial
herbaceous vegetation accounts for
pwtle 5 >80% vegetative cover 2006-2010
Composite variable of developed and
UINDEX 5 agricultural land cover No
pdev 5 All developed land cover No
Mixture of some constructed materials
but mostly vegetation; < 20%
pdevo 5 impervious surface No
Mixture of constructed materials and
vegetation; 20% to 49% impervious
pdevl 5 surface No
Mixture of constructed materials and
vegetation; 50% to 79% impervious
pdevm 5 surface No
Highly developed areas; 80% to 100%
pdevh 5 impervious surface No
Composite variable of pasture/hay and
pagr 5 cultivated crop land cover No
Notes
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Included as part of green space
composite variable
Does not meet definition of green
space
Does not meet definition of green
space
Included as part of green space
composite variable
Does not meet definition of green
space
Does not meet definition of green
space
Does not meet definition of green
space
Does not meet definition of green
space
B-ll
-------
Variable Variable Name Counties Variable Notes EQIVersion Notes
Grasses, legumes, or grass-legume
mixtures for livestock grazing;
Percentage of county land area production of seed or hay crops;
that is classified as pasture/hay pasture/hay vegetation accounts for Does not meet definition of green
land cover pagrp 5 >20% total vegetation No space
Production of annual crops; crop
Percentage of county land area vegetation accounts for >20% total
that is classified as cultivated crop vegetation; includes land being actively Does not meet definition of green
land cover pagrc 5 tilled No space
NATIONAL WALKABILITY INDEX (EPA)
NOTES: THE NATIONAL WALKABILITY INDEX IS A NATIONWIDE GEOGRAPHIC DATA RESOURCE THAT RANKS BLOCK GROUPS ACCORDING TO THEIR RELATIVE
WALKABILITY. THE NATIONAL DATASET INCLUDES WALKABILITY SCORES FOR ALL BLOCK GROUPS, AS WELL AS THE UNDERLYING ATTRIBUTES THAT ARE USED TO
RANK THE BLOCK GROUPS. DATAARE AVAILABLE FOR DOWNLOAD FROM THE EPASMARTGROWTH WEB SITE (HTTPSV/WWW.EPA.GOV/SMARTGROWTH/SMART-
LOCATION-MAPPING#WALKABILITYV
Variable Variable Name Counties Variable Notes EQI Version Notes
Scores were available at block group; county score
created by adding block group scores, then taking mean
National walkability of the block group scores based on county population
index score Sum_NWIBG 3143 proportions 2006-2010
B-12
-------
Appendix III: Changes in Variables
from EQI 2000-2005 to EQI 2006-2010
Table A: Variables Added
Domain Data Source Variable Variable Name Notes
Water Safe Drinking Water Information Total coliform, proportion Coliform_Sum Added to drinking water quality construct
System (SDWIS)
Land Mine Safety and Health Administration Primarily coal mines, mines per county Std_coal_prim_pop_ln Part of new mining activity construct
(MSHA) Mines Data Set population
Primarily metal mines, mines per county Std_coal_prim_pop_ln Part of new mining activity construct
population
Primarily nonmetal mines, mines per Std_coal_prim_pop_ln Part of new mining activity construct
county population
Primarily sand and gravel mines, mines Std_coal_prim_pop_ln Part of new mining activity construct
per county population
Primarily stone mines, mines per county Std_coal_prim_pop_ln Part of new mining activity construct
population
Sociodemographic United States Census Measure of income inequality GINI_est Added to socioeconomic construct
(proportion)
United States Department of Percent county employed in creative Num_CreatClass County creative typology construct
Agriculture Economic Research class
Service Creative Class
United States Election Atlas Percent county voting Democratic in DEM02008 County political valence construct
2008
Built TIGER Files Proportion of all roads that are SecondaryRoadProportion Replaced proportion primary road and
secondary roads highways
EnviroAtlas Land Cover Combined natural land cover and open NINDEX_open Green Space construct
space developed
National Walkability Index (EPA) National walkability index score Sum_NWIBG Walkability construct
Table B: Variables Changed
Domain Data Source Variable Variable Name Variable Replaced Variable Replaced Name
Sociodemographic United States Census Bachelor's degree or higher, Pct_BS Percent of persons with more Pct_hs_more
percent of persons age 25 years+ than a high school education
Table C: Variables Deleted
Percent of families in poverty Pct_Fam_Pov
Occupants per room ln_Occs_Room
Percent of persons less than Pct_pers_lt_pov
poverty level
Median number of rooms in Med_rooms
residence
Domain Data Source Variable Variable Name Reason Not Used
Land National Geochemical Mean level of arsenic from sampled county sources Mean_as_ln Data quality
Survey
Mean level of selenium from sampled county sources Mean_se_ln Data quality
Mean level of mercury from sampled county sources
Mean_hg_ln
Data quality
Mean level of lead from sampled county sources
Mean_pb_ln
Data quality
Mean level of zinc from sampled county sources
Mean_zn_ln
Data quality
Mean level of copper from sampled county sources
Mean_cu_ln
Data quality
Mean level of aluminum from sampled county sources
Mean_al_pct
Data quality
Mean level of sodium from sampled county sources
Mean_na_pct
Data quality
Mean level of magnesium from sampled county sources
Mean_mg_pct_ln
Data quality
C-l
-------
Table C: continued
Domain
Data Source
Built
Dun & Bradstreet
Built TIGER files
Sociodemographic United States Census
Variable
Mean level of titanium from sampled county sources
Mean level of calcium from sampled county sources
Mean level of manganese from sampled county sources
Mean level of iron from sampled county sources
Mean level of phosphorus from sampled county sources
Rate of transportation-related businesses per county
Rate of entertainment businesses per county
Proportion of all roads that are highways
Proportion of all roads that are primary roads
Percent of persons less than poverty level
Variable Name
Mean_ti_pct_ln
Mean_ca_pct_ln
Mean_mn
Mean_fe_pct_ln
mean_al_pct
rate_trans_env_log
rate_ent_env_log
hwyprop
primaryprop
pct_pers_lt_pov
Percent of persons who do not speak English pct_no_eng
Percent of persons with more than high school education pct_hs_more
Percent of persons who work outside their county of work_out_co
residence
Median number of rooms in residence med_rooms
Percent of residences with more than 10 units pct_mt_1 Ounitsjog
Reason Not Used
Data quality
Data quality
Data quality
Data quality
Data quality
Captured by public
transportation, commuting
times and roads
Dropped because there was
no clear association with
health
Both variables replaced with
secondary roads
Replaced with percent of
families below poverty level
Replaced with percent of
persons with a bachelor's
degree
Replaced with occupants
per room
Water Watershed Assessment, Sewage Permits per 1000 km of Stream in County SEWAGENPDESperKM Used group variable
Tracking and Environmental
Results Program Database/
REACH Address Database
Industrial Permits per 1000 km of stream in county
INDNPDESperKM
Used group variable
Stormwater Permits per 1000 km of stream in county
STORMNPDESperKM
Used group variable
Number of days closed per event in county 2002
numDays_Close_Activity_2002
Not enough counties
Number of days per contamination advisory event in numDays_Cont_Activity_2002 Not enouqh counties
county 2002
Number of days per rain advisory event in county 2002 numDays_Rain_Activity_2002 Not enough counties
Water National Atmospheric Magnesium (Mg) precipitation weighted mean (mg/L) Mgjn Correlated
Deposition Program
Sodium (Na) precipitation weighted mean (mg/L)
Najn
Correlated
Ammonium (NH4) precipitation weighted mean (mg/L)
NH4_mean
Correlated
/ater
National Contaminant
Occurrence Database
Beryllium - average
W_Be_ln (mg/L)
Zeros
Thallium - average
W_TI_ln (mg/L)
Correlated
Lindane - average
W_Lindane_ln (mg/L)
Correlated
Toxaphene - average
W_Toxaphene_ln (ug/L)
Correlated
Oxamyl (Vydate) - average
W_Oxamyl_ln (ug/L)
Correlated
Hexachlorocyclopentadiene - average
W_HCCPD_ln (ug/L)
Correlated
Carbofuran - average
W_Carbofuran_ln (ug/L)
Correlated
Alachlor - average
W_Alachlor_ln (ug/L)
Correlated
Heptachlor - average
W_Heptachlor_ln (ug/L)
Correlated
Heptachlor epoxide - average
W_Heptachlor_epox_ln (ug/L)
Correlated
2,4,5-TP (Silvex) - average
W_silvex_ln (ug/L)
Correlated
Hexachlorobenzene - average
WJ-ICBJn (ug/L)
Correlated
1,2,4-Trichlorobenzene - average
W_124TCIB_ln (ug/L)
Correlated
1,2-Dichlorobenzene (o-Dichlorobenzene) - average
W_ODCB_ln (ug/L)
Correlated
Vinyl chloride - average
W_VCM_ln (ug/L)
Correlated
C-2
-------
Table C: continued
Domain
Data Source
Variable
Variable Name
Reason Not Used
Carbon Tetrachloride - average
W_CCI4_ln (ug/L)
Correlated
1,1,2-Trichloroethane - average
W_112TCA_ln (ug/L)
Correlated
1,1 -Dichloroethylene - average
W_11 DCEJn (ug/L)
Correlated
trans-1,2-Dichloroethylene - average
W_t12DCE_ln (ug/L)
Correlated
1,2-Dichloroethane (Ethylene Dichloride) - average
W_EDC_ln (ug/L)
Correlated
1,2-Dichloropropane - average
W_PDC_ln (ug/L)
Correlated
Benzene - average
W_CI1benz_ln (ug/L)
Correlated
\ir
National-Scale Air Toxics
Assessment
2,4-toluene diisocyanate
A_TDI_ln
Correlated
2-chloroacetophenone
A_2Clacephen_ln
Correlated
2-nitropropane
A_2NP_ln
Correlated
4-nitrophenol
A_PNP_ln
Correlated
Acetonitrile
A_CH3CN_ln
Correlated
Acetophenone
A_Acetophenone_ln
Correlated
Acrolein
A_Aroclein_ln
Correlated
Acrylonitrile
A_C3H3N_ln
Correlated
Antimony compounds
A_Sb_ln
Correlated
Biphenyl
A_biphenyl_ln
Correlated
Bromoform
A_Bromoform_ln
Correlated
Cadmium compounds
A_Cd_ln
Correlated
Carbon disulfide
A_CS2_ln
Correlated
Carbon sulfide
A_CS_ln
Correlated
Cresol/cresylic acid
A_Cresol_ln
Correlated
Cumene
A_Cumene_ln
Correlated
Diesel engine emissions
A_Diesel_ln
Correlated
Dimethyl formamide
A_DMF_ln
Correlated
Dimethyl phthalates
A_Me2_phatalte_ln
Correlated
Dimethyl sulfate
A_Me2S04_ln
Correlated
Epichlorohydrin
A_ECH_ln
Correlated
Ethyl acrylate
A_Etacrylate_ln
Correlated
Ethylene glycol
A_EGLY_ln
Correlated
Ethylene oxide
A_EOx_ln
Correlated
Ethylidene dichloride
A_EdCI2_ln
Correlated
Hexachlorobenzene
A_HCB_ln
Correlated
Hexachlorobutadiene
A_HCBD_ln
Correlated
Hexachlorocyclopentadiene
A_HCCPD_ln
Correlated
Hexane
A_Hexane_ln
Correlated
Lead compounds
A_Pb_ln
Correlated
Mercury compounds
A_Hg_ln
Correlated
Methanol
A_MeOH_ln
Correlated
Methyl isobutyl ketone
A_MIBK_ln
Correlated
Methyl methacrylate
A_MMA_ln
Correlated
Methyl chloride
A_MeCI_ln
Correlated
Methylhydrazine
A_Mehydrazine_ln
Correlated
MTBE
A_MTBE_ln
Correlated
Nitrobenzene
A_nitrobenzene_ln
Correlated
N,N-dimethylaniline
A_DMA_ln
Correlated
o-toluidine
A_otoluidine_ln
Correlated
PAH/POM
A_PAHPOM_ln
Correlated
Pentachlorophenol
A_PCP_ln
Correlated
C-3
-------
Table C: continued
Domain Data Source Variable Variable Name Reason Not Used
Phosphorus A_P_ln Correlated
Propylene oxide A_ProO_ln Correlated
Selenium compounds
A_Se_ln
Correlated
Styrene
A_Styrene_ln
Correlated
Tetrachloroethylene
A_CI4C2_ln
Correlated
Toluene
A_Toluene_ln
Correlated
Triethylamine
A_Et3N_ln
Correlated
Vinyl acetate
A_VyAc_ln
Correlated
Vinylidene chloride
A_11DCE_ln
Correlated
C-4
-------
Appendix IV: Table of Highly Correlated
for Each Domain
Variables
Air Domain
Correlation
Variable Correlated Variable Coefficient Variable Used to Represent Group
1 -1 -1 -trichloroethane Methylene chloride 0.73 Methylene chloride
1-4-dichlorobenzene 0.70
Vinylidene chloride
2-2-4-trimethylpentane
2-chloroacetophenone
2-nitropropane
Ethylbenzene
2-2-4-trimethylpentane
Carbon disulfide
Cumene
Diesel engine emissions
Ethylene glycol
Hexane
Methanol
Methyl isobutyl ketone
MTBE
Naphthalene
Toluene
Xylenes
Ethylbenzene
Vinylidene chloride
4-4-methylenediphenyl diisocyanate
Acetophenone
Acrolein
Benzene
Biphenyl
1-3-butadiene
Tetrachloroethylene
Cresol cresylic acid
Cumene
Diesel engine emissions
Ethylene glycol
Triethylamine
Hexane
Mercury compounds
Dimethyl phthalate
Methanol
Methyl isobutyl ketone
Methyl methacrylate
MTBE
Naphthalene
Pahpom
4-nitrophenol
Propionaldehyde
Selenium compounds
Styrene
2-4-toluene diisocyanate
Toluene
Vinyl acetate
Xylenes
Benzyl chloride
Bromoform
Methylhydrazine
Chloroprene
Allyl chloride
n-n-dimethylaniline
2-4-dinitrotoluene
Nitrobenzene
o-toluidine
0.73
0.72
0.80
0.72
0.71
0.75
0.74
0.75
0.71
0.71
0.71
0.71
0.74
0.95
0.72
0.82
0.75
0.74
0.82
0.76
0.83
0.72
0.71
0.85
0.88
0.86
0.76
0.92
0.82
0.72
0.85
0.83
0.75
0.79
0.88
0.77
0.82
0.73
0.76
0.82
0.72
0.88
0.78
0.95
0.71
0.95
0.96
0.70
0.76
0.77
0.74
0.76
0.72
Ethylbenzene
Ethylbenzene
Benzyl chloride
Chloroprene
D-l
-------
Air Domain
Variable
4-4-methylenediphenyl diisocyanate
Acetophenone
Acrolein
Correlation
Correlated Variable
Coefficient
Ethylbenzene
0.83
2-2-4-trimethylpentane
0.82
Acetophenone
0.74
Acrolein
0.72
Benzene
0.73
Biphenyl
0.70
1-3-butadiene
0.76
Cumene
0.84
Diesel engine emissions
0.75
Ethylene glycol
0.86
Triethylamine
0.79
Hexane
0.82
Mercury compounds
0.76
Dimethyl phthalate
0.76
Methanol
0.83
Methyl isobutyl ketone
0.82
Methyl methacrylate
0.77
MTBE
0.74
Naphthalene
0.79
Pahpom
0.72
Phenol
0.71
4-nitrophenol
0.78
Selenium compounds
0.71
Styrene
0.79
2-4-toluene diisocyanate
0.75
Toluene
0.77
Vinyl acetate
0.80
Xylenes
0.84
Ethylbenzene
0.76
2-2-4-trimethylpentane
0.75
4-4-methylenediphenyl diisocyanate
0.74
Biphenyl
0.78
1-3-butadiene
0.72
Cresol cresylic acid
0.76
Cumene
0.78
Ethylene glycol
0.78
Triethylamine
0.71
Hexane
0.74
Mercury compounds
0.75
Methanol
0.78
Methyl isobutyl ketone
0.76
MTBE
0.74
Naphthalene
0.76
Pahpom
0.76
Phenol
0.73
4-nitrophenol
0.81
Selenium compounds
0.72
Toluene
0.70
Vinyl acetate
0.70
Xylenes
0.77
Ethylbenzene
0.77
2-2-4-trimethylpentane
0.74
4-4-methylenediphenyl diisocyanate
0.72
1-3-butadiene
0.74
Cresol cresylic acid
0.81
Cumene
0.74
Ethylene glycol
0.76
Hexane
0.73
Methanol
0.76
Methyl isobutyl ketone
0.75
MTBE
0.73
Naphthalene
0.75
Pahpom
0.71
Propionaldehyde
0.75
Xylenes
0.77
Variable Used to Represent Group
Ethylbenzene
Ethylbenzene
Ethylbenzene
D-2
-------
Air Domain
Variable
Allyl chloride
Arsenic compounds
Benzene
Biphenyl
Bromoform
Correlation
Correlated Variable
Coefficient
Chloroprene
0.90
2-nitropropane
0.76
Acetonitrile
0.81
n-n-dimethylaniline
0.96
Epichlorohydrin
0.85
Ethyl acrylate
0.78
Hexachlorobutadiene
0.73
Hexachlorocyclopentadiene
0.70
Nitrobenzene
0.96
o-toluidine
0.85
Propylene oxide
0.77
1-2-4-trichlorobenzene
0.78
Chromium compounds
0.80
Cadmium compounds
0.80
Lead compounds
0.74
Ethylbenzene
0.85
2-2-4-trimethylpentane
0.82
4-4-methylenediphenyl diisocyanate
0.73
1-3-butadiene
0.90
Tetrachloroethylene
0.85
Cumene
0.77
Diesel engine emissions
0.76
Ethylene glycol
0.80
Hexane
0.81
Mercury compounds
0.71
Methanol
0.79
Methyl isobutyl ketone
0.74
MTBE
0.70
Naphthalene
0.80
4-nitrophenol
0.74
Styrene
0.70
Toluene
0.96
Xylenes
0.85
Ethylbenzene
0.75
2-2-4-trimethylpentane
0.76
4-4-methylenediphenyl diisocyanate
0.70
Acetophenone
0.78
1-3-butadiene
0.70
Cresol cresylic acid
0.74
Cumene
0.77
Ethylene glycol
0.77
Hexane
0.73
Mercury compounds
0.76
Methanol
0.77
Methyl isobutyl ketone
0.74
MTBE
0.71
Naphthalene
0.77
Pahpom
0.80
Phenol
0.74
4-nitrophenol
0.74
Selenium compounds
0.72
Toluene
0.71
Xylenes
0.76
Benzyl chloride
0.70
Methylhydrazine
0.94
Variable Used to Represent Group
Chloroprene
Chromium compounds
Ethylbenzene
Ethylbenzene
Benzyl chloride
D-3
-------
Air Domain
Correlation
Variable
Correlated Variable
Coefficient
Variable Used to Represent Group
1 -3-butadiene
Ethylbenzene
0.84
Ethylbenzene
2-2-4-trimethylpentane
0.83
4-4-methylenediphenyl diisocyanate
0.76
Acetophenone
0.72
Acrolein
0.74
Benzene
0.90
Biphenyl
0.70
Tetrachloroethylene
0.74
Cresol cresylic acid
0.71
Cumene
0.80
Diesel engine emissions
0.72
Ethylene glycol
0.83
Triethylamine
0.74
Hexane
0.81
Mercury compounds
0.76
Methanol
0.81
Methyl isobutyl ketone
0.79
Methyl methacrylate
0.71
MTBE
0.73
Naphthalene
0.81
Pahpom
0.72
4-nitrophenol
0.77
Selenium compounds
0.70
Styrene
0.73
2-4-toluene diisocyanate
0.70
Toluene
0.94
Vinyl acetate
0.73
Xylenes
0.84
Acrylonitrile
Trichloroethylene
0.74
Trichloroethylene
Cadmium compounds
Chromium compounds
0.71
Chromium compounds
Arsenic compounds
0.80
Acetonitrile
Chloroprene
0.80
Chloroprene
Allyl chloride
0.81
n-n-dimethylaniline
0.80
2-4-dinitrotoluene
0.75
Epichlorohydrin
0.76
Nitrobenzene
0.79
o-toluidine
0.75
Propylene oxide
0.77
Tetrachloroethylene
Ethylbenzene
0.72
Ethylbenzene
2-2-4-trimethylpentane
0.72
Benzene
0.85
1-3-butadiene
0.74
Naphthalene
0.73
Toluene
0.82
Xylenes
0.72
Cresol cresylic acid
Ethylbenzene
0.77
Ethylbenzene
2-2-4-trimethylpentane
0.71
Acetophenone
0.76
Acrolein
0.81
Biphenyl
0.74
1-3-butadiene
0.71
Cumene
0.73
Ethylene glycol
0.75
Triethylamine
0.71
Mercury compounds
0.73
Methanol
0.74
Methyl isobutyl ketone
0.75
Naphthalene
0.78
Pahpom
0.76
Phenol
0.75
Propionaldehyde
0.71
Xylenes
0.78
D-4
-------
Air Domain
Variable
Carbon disulfide
Cumene
1-4-dichlorobenzene
Diesel engine emissions
Correlated Variable
Ethylbenzene
Vinylidene chloride
Cumene
Ethylene glycol
Methanol
Methyl isobutyl ketone
Xylenes
Ethylbenzene
Vinylidene chloride
2-2-4-trimethylpentane
4-4-methylenediphenyl diisocyanate
Acetophenone
Acrolein
Benzene
Biphenyl
1-3-butadiene
Cresol cresylic acid
Carbon disulfide
Diesel engine emissions
Ethylene glycol
Triethylamine
Hexane
Mercury compounds
Dimethyl phthalate
Methanol
Methyl isobutyl ketone
Methyl methacrylate
MTBE
Naphthalene
Pahpom
Phenol
4-nitrophenol
Selenium compounds
Styrene
2-4-toluene diisocyanate
Toluene
Vinyl acetate
Xylenes
Methylene chloride
1-1-1 -trichloroethane
Ethylbenzene
Vinylidene chloride
2-2-4-trimethylpentane
4-4-methylenediphenyl diisocyanate
Benzene
1-3-butadiene
Cumene
Ethylene glycol
Triethylamine
Hexane
Mercury compounds
Methanol
Methyl isobutyl ketone
MTBE
Naphthalene
4-nitrophenol
Selenium compounds
Styrene
2-4-toluene diisocyanate
Toluene
Vinyl acetate
Xylenes
Correlation
Coefficient
0.72
0.80
0.70
0.74
0.74
0.73
0.72
0.87
0.72
0.85
0.84
0.78
0.74
0.77
0.77
0.80
0.73
0.70
0.77
0.89
0.82
0.88
0.81
0.74
0.88
0.86
0.76
0.83
0.84
0.79
0.78
0.81
0.76
0.81
0.77
0.81
0.80
0.88
0.80
0.70
0.86
0.71
0.88
0.75
0.76
0.72
0.77
0.78
0.70
0.85
0.75
0.78
0.74
0.73
0.78
0.74
0.71
0.74
0.71
0.78
0.72
0.85
Variable Used to Represent Group
Ethylbenzene
Ethylbenzene
Methylene chloride
Ethylbenzene
D-5
-------
Air Domain
Correlation
Variable
Correlated Variable
Coefficient
Variable Used to Represent Group
n-n-di methylanil ine
Dimethyl formamide
2-4-dinitrotoluene
Epichlorohydrin
Ethylidene dichloride
Chloroprene
0.92
2-nitropropane
0.77
Allyl chloride
0.96
Acetonitrile
0.80
2-4-dinitrotoluene
0.92
Epichlorohydrin
0.86
Ethyl acrylate
0.77
Hexachlorobutadiene
0.72
Hexachlorocyclopentadiene
0.72
Nitrobenzene
0.95
o-toluidine
0.86
Propylene oxide
0.78
1-2-4-trichlorobenzene
0.78
Ethyl chloride
0.71
Chloroprene
0.88
2-nitropropane
0.74
Allyl chloride
0.89
A_CH3CN
0.75
n-n-dimethylaniline
0.92
Epichlorohydrin
0.84
Ethyl acrylate
0.76
Hexachlorocyclopentadiene
0.70
Nitrobenzene
0.88
o-toluidine
0.86
Propylene oxide
0.70
1-2-4-trichlorobenzene
0.76
Chloroprene
0.84
Allyl chloride
0.85
Acetonitrile
0.76
n-n-dimethylaniline
0.86
2-4-dinitrotoluene
0.84
Ethyl acrylate
0.77
Nitrobenzene
0.81
o-toluidine
0.80
Propylene oxide
0.75
1-2-4-trichlorobenzene
0.74
Vinyl chloride
0.82
Chloroprene
Ethyl chloride
Chloroprene
Chloroprene
Vinyl chloride
D-6
-------
Air Domain
Variable
Ethylene glycol
Ethylene oxide
Triethylamine
Ethyl acrylate
Hexachlorobenzene
Correlation
Correlated Variable
Coefficient
Ethylbenzene
0.88
Vinylidene chloride
0.75
2-2-4-trimethylpentane
0.86
4-4-methylenediphenyl diisocyanate
0.86
Acetophenone
0.78
Acrolein
0.76
Benzene
0.80
Biphenyl
0.77
1-3-butadiene
0.83
Cresol cresylic acid
0.75
Carbon disulfide
0.74
Cumene
0.89
Diesel engine emissions
0.78
Triethylamine
0.83
Hexane
0.87
Mercury compounds
0.84
Dimethyl phthalate
0.76
Methanol
0.93
Methyl isobutyl ketone
0.91
Methyl methacrylate
0.79
MTBE
0.81
Naphthalene
0.86
Pahpom
0.78
Phenol
0.75
4-nitrophenol
0.83
Propionaldehyde
0.73
Selenium compounds
0.78
Styrene
0.81
2-4-toluene diisocyanate
0.78
Toluene
0.84
Vinyl acetate
0.82
Xylenes
0.90
Ethylene dichloride
0.72
Ethylbenzene
0.79
2-2-4-trimethylpentane
0.76
4-4-methylenediphenyl diisocyanate
0.79
Acetophenone
0.71
1-3-butadiene
0.74
Cresol cresylic acid
0.71
Cumene
0.82
Diesel engine emissions
0.70
Ethylene glycol
0.83
Hexane
0.79
Mercury compounds
0.75
Methanol
0.80
Methyl isobutyl ketone
0.81
Methyl methacrylate
0.72
MTBE
0.70
Naphthalene
0.80
Pahpom
0.70
4-nitrophenol
0.71
Styrene
0.73
2-4-toluene diisocyanate
0.74
Toluene
0.74
Vinyl acetate
0.77
Xylenes
0.81
Chloroprene
0.80
Allyl chloride
0.78
n-n-dimethylaniline
0.77
2-4-dinitrotoluene
0.76
Epichlorohydrin
0.77
Nitrobenzene
0.75
o-toluidine
0.76
Polychlorinated biphenyls
0.83
Variable Used to Represent Group
Ethylbenzene
Ethylene dichloride
Ethylbenzene
Chloroprene
Polychlorinated biphenyls
D-7
-------
Air Domain
Variable
Hexachlorobutadiene
Hexachlorocyclopentadiene
Hexane
Hydrogen fluoride
Mercury compounds
Correlation
Correlated Variable
Coefficient
Chloroprene
0.70
Allyl chloride
0.73
n-n-dimethylaniline
0.72
Hexachlorocyclopentadiene
0.93
Nitrobenzene
0.73
Chloroprene
0.71
Allyl chloride
0.70
n-n-dimethylaniline
0.72
2-4-dinitrotoluene
0.70
Hexachlorobutadiene
0.93
Ethylbenzene
0.92
Vinylidene chloride
0.74
2-2-4-trimethylpentane
0.92
4-4-methylenediphenyl diisocyanate
0.82
Acetophenone
0.74
Acrolein
0.73
Benzene
0.81
Biphenyl
0.73
1-3-butadiene
0.81
Cumene
0.88
Diesel engine emissions
0.85
Ethylene glycol
0.87
Triethylamine
0.79
Mercury compounds
0.80
Dimethyl phthalate
0.72
Methanol
0.87
Methyl isobutyl ketone
0.83
Methyl methacrylate
0.76
MTBE
0.81
Naphthalene
0.86
Pahpom
0.73
4-nitrophenol
0.79
Selenium compounds
0.72
Styrene
0.80
2-4-toluene diisocyanate
0.77
Toluene
0.85
Vinyl acetate
0.79
Xylenes
0.92
Hydrochloric acid
0.91
Ethylbenzene
0.82
2-2-4-trimethylpentane
0.82
4-4-methylenediphenyl diisocyanate
0.76
Acetophenone
0.75
Benzene
0.71
Biphenyl
0.76
1-3-butadiene
0.76
Cresol cresylic acid
0.73
Cumene
0.81
Diesel engine emissions
0.75
Ethylene glycol
0.84
Triethylamine
0.75
Hexane
0.80
Methanol
0.82
Methyl isobutyl ketone
0.81
Methyl methacrylate
0.72
MTBE
0.74
Naphthalene
0.84
Pahpom
0.75
Phenol
0.72
4-nitrophenol
0.80
Propionaldehyde
0.73
Selenium compounds
0.91
Styrene
0.74
Toluene
0.76
Vinyl acetate
0.76
Xylenes
0.82
Variable Used to Represent Group
Chloroprene
Chloroprene
Ethylbenzene
Hydrochloric acid
Ethylbenzene
D-8
-------
Air Domain
Variable
Dimethyl phthalate
Dimethyl sulfate
Methyl chloride
Methylhydrazine
Methanol
Correlation
Correlated Variable
Coefficient
Ethylbenzene
0.73
2-2-4-trimethylpentane
0.72
4-4-methylenediphenyl diisocyanate
0.76
Cumene
0.74
Ethylene glycol
0.76
Hexane
0.72
Methanol
0.74
Methyl isobutyl ketone
0.73
Methyl methacrylate
0.76
Naphthalene
0.71
Styrene
0.75
Xylenes
0.74
Benzyl chloride
0.90
Carbon tetrachloride
0.94
Benzyl chloride
0.71
2-chloroacetophenone
0.96
Bromoform
0.94
Ethylbenzene
0.88
Vinylidene chloride
0.75
2-2-4-trimethylpentane
0.85
4-4-methylenediphenyl diisocyanate
0.83
Acetophenone
0.78
Acrolein
0.76
Benzene
0.79
Biphenyl
0.77
1-3-butadiene
0.81
Cresol cresylic acid
0.74
Carbon disulfide
0.74
Cumene
0.88
Diesel engine emissions
0.78
Ethylene glycol
0.93
Triethylamine
0.80
Hexane
0.87
Mercury compounds
0.82
Dimethyl phthalate
0.74
Methyl isobutyl ketone
0.89
Methyl methacrylate
0.78
MTBE
0.82
Naphthalene
0.84
Pahpom
0.78
Phenol
0.76
4-nitrophenol
0.82
Propionaldehyde
0.72
Selenium compounds
0.77
Styrene
0.81
2-4-toluene diisocyanate
0.76
Toluene
0.82
Vinyl acetate
0.79
Xylenes
0.89
Variable Used to Represent Group
Ethylbenzene
Benzyl chloride
Carbon tetrachloride
Benzyl chloride
Ethylbenzene
D-9
-------
Air Domain
Variable
Methyl isobutyl ketone
Methyl methacrylate
Correlation
Correlated Variable
Coefficient
Ethylbenzene
0.86
Vinylidene chloride
0.71
2-2-4-trimethylpentane
0.83
4-4-methylenediphenyl diisocyanate
0.82
Acetophenone
0.76
Acrolein
0.75
Benzene
0.74
Biphenyl
0.74
1-3-butadiene
0.79
Cresol cresylic acid
0.75
Carbon disulfide
0.73
Cumene
0.86
Diesel engine emissions
0.74
Ethylene glycol
0.91
Triethylamine
0.81
Hexane
0.83
Mercury compounds
0.81
Dimethyl phthalate
0.73
Methanol
0.89
Methyl methacrylate
0.77
MTBE
0.81
Naphthalene
0.82
Pahpom
0.77
Phenol
0.78
4-nitrophenol
0.79
Selenium compounds
0.76
Styrene
0.81
2-4-toluene diisocyanate
0.77
Toluene
0.79
Vinyl acetate
0.76
Xylenes
0.89
Ethylbenzene
0.77
2-2-4-trimethylpentane
0.75
4-4-methylenediphenyl diisocyanate
0.77
1-3-butadiene
0.71
Cumene
0.76
Ethylene glycol
0.79
Triethylamine
0.72
Hexane
0.76
Mercury compounds
0.72
Dimethyl phthalate
0.76
Methanol
0.78
Methyl isobutyl ketone
0.77
Naphthalene
0.74
4-nitrophenol
0.72
Styrene
0.83
Toluene
0.71
Vinyl acetate
0.72
Xylenes
0.78
Variable Used to Represent Group
Ethylbenzene
Ethylbenzene
D-10
-------
Air Domain
Correlation
Variable
Correlated Variable
Coefficient
Variable Used to Represent Group
MTBE
Naphthalene
Nickel compounds
Ethylbenzene
0.79
Vinylidene chloride
0.71
2-2-4-trimethylpentane
0.79
4-4-methylenediphenyl diisocyanate
0.74
Acetophenone
0.74
Acrolein
0.73
Benzene
0.70
Biphenyl
0.71
1-3-butadiene
0.73
Cumene
0.83
Diesel engine emissions
0.73
Ethylene glycol
0.81
Triethylamine
0.70
Hexane
0.81
Mercury compounds
0.74
Methanol
0.82
Methyl isobutyl ketone
0.81
Naphthalene
0.78
Pahpom
0.71
Phenol
0.71
4-nitrophenol
0.74
Selenium compounds
0.70
Styrene
0.72
2-4-toluene diisocyanate
0.73
Toluene
0.73
Xylenes
0.79
Ethylbenzene
0.87
Vinylidene chloride
0.71
2-2-4-trimethylpentane
0.88
4-4-methylenediphenyl diisocyanate
0.79
Acetophenone
0.76
Acrolein
0.75
Benzene
0.80
Biphenyl
0.77
1-3-butadiene
0.81
Tetrachloroethylene
0.73
Cresol cresylic acid
0.78
Cumene
0.84
Diesel engine emissions
0.78
Ethylene glycol
0.86
Triethylamine
0.80
Hexane
0.86
Mercury compounds
0.84
Dimethyl phthalate
0.71
Methanol
0.84
Methyl isobutyl ketone
0.82
Methyl methacrylate
0.74
MTBE
0.78
Pahpom
0.84
Phenol
0.73
4-nitrophenol
0.79
Propionaldehyde
0.74
Selenium compounds
0.77
Styrene
0.76
2-4-toluene diisocyanate
0.70
Toluene
0.83
Vinyl acetate
0.78
Xylenes
0.88
Chromium compounds
0.79
Ethylbenzene
Ethylbenzene
Chromium compounds
D-ll
-------
Air Domain
Variable
Nitrobenzene
o-toluidine
Pahpom
Lead compounds
Phenol
Correlation
Correlated Variable
Coefficient
Chloroprene
0.88
2-nitropropane
0.76
Allyl chloride
0.96
Acetonitrile
0.79
n-n-dimethylaniline
0.95
2-4-dinitrotoluene
0.88
Epichlorohydrin
0.81
Ethyl acrylate
0.75
Hexachlorobutadiene
0.70
o-toluidine
0.82
Propylene oxide
0.77
1-2-4-trichlorobenzene
0.76
Chloroprene
0.84
2-nitropropane
0.72
Allyl chloride
0.85
Acetonitrile
0.75
n-n-dimethylaniline
0.86
2-4-dinitrotoluene
0.86
Epichlorohydrin
0.80
Ethyl acrylate
0.76
Nitrobenzene
0.82
Propylene oxide
0.77
1-2-4-trichlorobenzene
0.76
Ethylbenzene
0.76
2-2-4-trimethylpentane
0.77
4-4-methylenediphenyl diisocyanate
0.72
Acetophenone
0.76
Acrolein
0.71
Biphenyl
0.80
1-3-butadiene
0.72
Cresol cresylic acid
0.76
Cumene
0.79
Ethylene glycol
0.78
Triethylamine
0.70
Hexane
0.73
Mercury compounds
0.75
Methanol
0.78
Methyl isobutyl ketone
0.77
MTBE
0.71
Naphthalene
0.84
Phenol
0.79
4-nitrophenol
0.76
Selenium compounds
0.72
Styrene
0.73
Xylenes
0.78
Chromium compounds
0.74
Arsenic compounds
0.74
Ethylbenzene
0.71
4-4-methylenediphenyl diisocyanate
0.71
Acetophenone
0.73
Biphenyl
0.74
Cresol cresylic acid
0.75
Cumene
0.78
Ethylene glycol
0.75
Mercury compounds
0.72
Methanol
0.76
Methyl isobutyl ketone
0.78
MTBE
0.71
Naphthalene
0.73
Pahpom
0.79
Styrene
0.74
Xylenes
0.72
Variable Used to Represent Group
Chloroprene
Chloroprene
Ethylbenzene
Chromium compounds
Ethylbenzene
D-12
-------
Air Domain
Correlation
Variable
Correlated Variable
Coefficient
Variable Used to Represent Group
4-nitrophenol
Ethylbenzene
0.81
Ethylbenzene
2-2-4-trimethylpentane
0.82
4-4-methylenediphenyl diisocyanate
0.78
Acetophenone
0.81
Benzene
0.74
Biphenyl
0.74
1-3-butadiene
0.77
Cumene
0.81
Diesel engine emissions
0.74
Ethylene glycol
0.83
Triethylamine
0.71
Hexane
0.79
Mercury compounds
0.80
Methanol
0.82
Methyl isobutyl ketone
0.79
Methyl methacrylate
0.72
MTBE
0.74
Naphthalene
0.79
Pahpom
0.76
Propionaldehyde
0.71
Selenium compounds
0.75
Styrene
0.75
2-4-toluene diisocyanate
0.70
Toluene
0.77
Vinyl acetate
0.73
Xylenes
0.81
Propylene oxide
Chloroprene
0.75
Chloroprene
Allyl chloride
0.77
Acetonitrile
0.77
n-n-dimethylaniline
0.78
2-4-dinitrotoluene
0.70
Epichlorohydrin
0.75
Nitrobenzene
0.77
o-toluidine
0.73
Propionaldehyde
Ethylbenzene
0.74
Ethylbenzene
2-2-4-trimethylpentane
0.73
Acrolein
0.75
Cresol cresylic acid
0.71
Ethylene glycol
0.73
Mercury compounds
0.73
Methanol
0.72
Naphthalene
0.74
4-nitrophenol
0.71
Selenium compounds
0.70
Xylenes
0.73
Selenium compounds
Ethylbenzene
0.76
Ethylbenzene
2-2-4-trimethylpentane
0.76
4-4-methylenediphenyl diisocyanate
0.71
Acetophenone
0.72
Biphenyl
0.72
1-3-butadiene
0.70
Cumene
0.76
Diesel engine emissions
0.71
Ethylene glycol
0.78
Hexane
0.72
Mercury compounds
0.91
Methanol
0.77
Methyl isobutyl ketone
0.76
MTBE
0.70
Naphthalene
0.77
Pahpom
0.72
4-nitrophenol
0.75
Propionaldehyde
0.70
Xylenes
0.77
D-13
-------
Air Domain
Variable
Styrene
1-2-4-trichlorobenzene
2-4-toluene diisocyanate
Correlation
Correlated Variable
Coefficient
Ethylbenzene
0.82
2-2-4-trimethylpentane
0.82
4-4-methylenediphenyl diisocyanate
0.79
Benzene
0.70
1 -3-butadiene
0.73
Cumene
0.81
Diesel engine emissions
0.74
Ethylene glycol
0.81
Triethylamine
0.73
Hexane
0.80
Mercury compounds
0.74
Dimethyl phthalate
0.75
Methanol
0.81
Methyl isobutyl ketone
0.81
Methyl methacrylate
0.83
MTBE
0.72
Naphthalene
0.76
Pahpom
0.73
Phenol
0.74
4-nitrophenol
0.75
Toluene
0.74
Vinyl acetate
0.73
Xylenes
0.83
Chloroprene
0.70
Allyl chloride
0.78
n-n-dimethylaniline
0.78
2-4-dinitrotoluene
0.76
Epichlorohydrin
0.74
Nitrobenzene
0.76
o-toluidine
0.74
Ethylbenzene
0.77
2-2-4-trimethylpentane
0.72
4-4-methylenediphenyl diisocyanate
0.75
1-3-butadiene
0.70
Cumene
0.77
Diesel engine emissions
0.71
Ethylene glycol
0.78
Triethylamine
0.74
Hexane
0.77
Methanol
0.76
Methyl isobutyl ketone
0.77
MTBE
0.73
Naphthalene
0.70
4-nitrophenol
0.70
Toluene
0.71
Vinyl acetate
0.70
Xylenes
0.77
Variable Used to Represent Group
Ethylbenzene
Chloroprene
Ethylbenzene
D-14
-------
Air Domain
Variable
Toluene
Vinyl acetate
Correlation
Correlated Variable
Coefficient
Ethylbenzene
0.88
Vinylidene chloride
0.71
2-2-4-trimethylpentane
0.88
4-4-methylenediphenyl diisocyanate
0.77
Acetophenone
0.70
Benzene
0.96
Biphenyl
0.71
1-3-butadiene
0.94
Tetrachloroethylene
0.82
Cumene
0.81
Diesel engine emissions
0.78
Ethylene glycol
0.84
Triethylamine
0.74
Hexane
0.85
Mercury compounds
0.76
Methanol
0.82
Methyl isobutyl ketone
0.79
Methyl methacrylate
0.71
MTBE
0.73
Naphthalene
0.83
4-nitrophenol
0.77
Styrene
0.74
2-4-toluene diisocyanate
0.71
Vinyl acetate
0.73
Xylenes
0.88
Ethylbenzene
0.79
2-2-4-trimethylpentane
0.78
4-4-methylenediphenyl diisocyanate
0.80
Acetophenone
0.70
1-3-butadiene
0.73
Cumene
0.80
Diesel engine emissions
0.72
Ethylene glycol
0.82
Triethylamine
0.77
Hexane
0.79
Mercury compounds
0.76
Methanol
0.79
Methyl isobutyl ketone
0.76
Methyl methacrylate
0.72
Naphthalene
0.78
4-nitrophenol
0.73
Styrene
0.73
2-4-toluene diisocyanate
0.70
Toluene
0.73
Xylenes
0.88
Variable Used to Represent Group
Ethylbenzene
Ethylbenzene
D-15
-------
Air Domain
Variable
Xylenes
Correlation
Correlated Variable
Coefficient
Ethylbenzene
0.99
Vinylidene chloride
0.74
2-2-4-trimethylpentane
0.95
4-4-methylenediphenyl diisocyanate
0.84
Acetophenone
0.77
Acrolein
0.77
Benzene
0.85
Biphenyl
0.76
1-3-butadiene
0.84
Tetrachloroethylene
0.72
Cresol cresylic acid
0.78
Carbon disulfide
0.72
Cumene
0.88
Diesel engine emissions
0.85
Ethylene glycol
0.90
Triethylamine
0.81
Hexane
0.92
Mercury compounds
0.82
Dimethyl phthalate
0.74
Methanol
0.89
Methyl isobutyl ketone
0.89
Methyl methacrylate
0.78
MTBE
0.79
Naphthalene
0.88
Pahpom
0.78
Phenol
0.72
4-nitrophenol
0.81
Propionaldehyde
0.73
Selenium compounds
0.77
Styrene
0.83
2-4-toluene diisocyanate
0.77
Toluene
0.88
Vinyl acetate
0.80
Variable Used to Represent Group
Ethylbenzene
Water Domain
Variable
Percent of county abnormally dry
Percent of county drought - moderate
Percent of county drought - severe
Percent of county drought - exceptional
Lindane - average
Thallium - average
Toxaphene - average
Oxamyl (Vydate) - average
Alachlor - average
2,4,5-TP (Silvex) - average
Hexachlorocyclopentadiene - average
Carbofuran - average
Correlation
Correlated Variable(s) Coefficient
Percent of county without drought, 0.94
Percent of county drought - moderate, 0.94
Percent of county drought - severe, 0.86
Percent of county drought - extreme 0.71
Percent of county without drought, 0.94
Percent of county abnormally dry, 0.94
Percent of county drought - severe, 0.86
Percent of county drought - extreme 0.71
Percent of county without drought, 0.86
Percent of county abnormally dry, Percent of county 0.86
drought - moderate, 0.94
Percent of county drought - extreme 0.71
Percent of county drought - moderate, 0.94
Percent of county drought - severe, 0.86
Percent of county drought - extreme 0.80
Barium - average 0.75
Cadmium - average 0.76
Endrin - average 0.80
Dalapon - average 0.70
Simazine - average 0.72
Picloram - average 0.73
Ethylene dibromide (EDB) - average 0.80
Chlordane - average 0.79
Variable Used To Represent Group
Percent of county drought - extreme
Percent of county drought - extreme
Percent of county drought - extreme
Percent of county drought - extreme
Barium - average
Cadmium - average
Endrin - average
Dalapon - average
Simazine - average
Picloram - average
Ethylene dibromide (EDB) - average
Chlordane - average
D-16
-------
Water Domain
Variable
Correlated Variable(s)
Correlation
Coefficient Variable Used To Represent Group
Heptachlor - average
Di(2-ethylhexyl) phthalate (DEHP) - average
Hexachlorobenzene - average Heptachlor - average
0.77
0.70
0.81
Di(2-ethylhexyl) phthalate (DEHP) - average
Heptachlor Epoxide - average
Di(2-ethylhexyl) phthalate (DEHP) - average
Hexachlorobenzene - average Heptachlor - average
0.73
0.74
0.81
Di(2-ethylhexyl) phthalate (DEHP) - average
Hexachlorobenzene - average
Di(2-ethylhexyl) phthalate (DEHP) - average
Heptachlor - average
Heptachlor Epoxide - average
0.77
0.70
0.74
Di(2-ethylhexyl) phthalate (DEHP) - average
1,2,4-Trichlorobenzene - average
Ethylbenzene - average
Vinyl chloride - average
Benzene - average
0.77
0.71
0.82
Ethylbenzene - average
1,2-Dichlorobenzene (o-Dichlorobenzene) 1,2,4-Trichlorobenzene - detect Ethylbenzene - average
- average Benzene - average
0.80
0.77
0.88
Ethylbenzene - average
Vinyl chloride - average
1,2-Dichlorobenzene (o-Dichlorobenzene) - average
1,2,4-Trichlorobenzene - detect
Ethylbenzene - average
Benzene - average
0.73
0.80
0.77
0.82
Ethylbenzene - average
Benzene - average
1,2-Dichlorobenzene (o-Dichlorobenzene) - average
1,2,4-Trichlorobenzene - detect
Ethylbenzene - average
Vinyl chloride - average
0.88
0.82
0.72
0.82
Ethylbenzene - average
1,1 -Dichloroethylene - average
cis1,2-Dichloroethylene - average Dichloroethylene - average
cis-1,2-Dichloroethylene - average
0.70
0.70
0.81
cis-1,2-Dichloroethylene - average
W_t12DCE_ln
cis-1,2-Dichloroethylene - average
1,1-Dichloroethylene - average cis-1,2-Dichloroethylene -
average
0.82
0.70
0.75
cis-1,2-Dichloroethylene - average
cis-1,2-Dichloroethylene - average
cis-1,2-Dichloroethylene - average
1,1 -Dichloroethylene - average Dichloroethylene - average
0.82
0.81
0.75
cis-1,2-Dichloroethylene - average
Carbon Tetrachloride - average
1,1,1 -Trichloroethane - average
0.71
1,1,1 -Trichloroethane - average
1,2-Dichloropropane - average
1,4-Dichlorobenzene (p-Dichlorobenzene) - average
0.72
1,4-Dichlorobenzene (p-Dichlorobenzene) -
average
1,1,2-Trichloroethane - average
Tetrachloroethylene - average
0.80
Tetrachloroethylene - average
Land Domain
Variable
Correlated Variable(s) Correlation Coefficient
Variable Used To Represent Group
Mean manganese
Mean iron percent 0.90
Mean iron percent
Percent weed acres
Percent harvested acres, 0.96
percent lime acres 0.95
Percent harvested acres
Percent lime acres
Percent harvested acres, 0.97
percent weed acres 0.95
Percent harvested acres
Sociodemographic Domain
Variable
Correlated Variable(s) Correlation Coefficient
Variable Used To Represent Group
Property crime rate
Violent crime rate 0.91
Violent crime rate
Built Domain
Variable Correlated Variable(s) Correlation Coefficient Variable Used To Represent Group
Secondary road proportion Street proportion 0.94 Street proportion
D-17
-------
-------
Appendix V: Sociodemographic and Built-Domain
Valence Correction
-0.1269
0.0161
0.1269
Percent unemployed
Harmful
Yes
-0.1979
0.0392
0.1979
Percent vacant
housing
Harmful
Yes
0.3824
0.1462
-0.3824
Household income
Beneficial
Yes
0.1458
0.0213
-0.1458
Percent renter-
occupied housing
Harmful
Yes
Yes
0.4833
0.2336
-0.4833
Percent creative
class
Beneficial
Yes
-0.0118
0.0001
0.0118
GINI
Harmful
Yes
Violent crime
Harmful
0.0234
Yes
Yes
0.0005
-0.0234
Percent Democratic
Beneficial
0.211
Yes
0.0445
-0.211
Count of occupants
per room
Harmful
-0.1085
Yes
0.0118
0.1085
Percent families less
than poverty level
Harmful
-0.298
Yes
0.298
Median household
value
Beneficial
0.4331
Yes
0.1876
-0.4331
Percent bachelor's
degree
Sociodemographic Overall
Loading
A priori Variable (Expected
Characteristic Sign)
Beneficial
Loading
(Actual)
0.4585
Match
(Expected
versus
Observed)
Necessary To
Multiply Vector of
Loadings by-1? (Loading)A2
Yes 0.2102
Modified Loadings
-0.4585
coefficient
-0.1625
0.0264
0.1625
Percent unemployed
Harmful
Yes
-0.2306
0.0532
0.2306
Percent vacant housing
Harmful
Yes
0.3700
0.1369
-0.3700
Household income
Beneficial
Yes
0.1827
0.0334
-0.1827
Percent renter-occupied housing
Harmful
Yes
Yes
0.4668
0.2179
-0.4668
Percent creative class
Beneficial
Yes
0.1162
0.0135
-0.1162
GINI
Harmful
Yes
Yes
Count of occupants per room
Harmful
-0.0055
Yes
0.0000
0.0055
Violent crime
Harmful
0.0094
Yes
Yes
0.0001
-0.0094
Median household value
Beneficial
0.4034
Yes
0.1627
-0.4034
Percent Democratic
Beneficial
0.2625
Yes
-0.2625
Percent families less than poverty
level
Harmful
-0.2591
Yes
0.0671
0.2591
Percent bachelor's degree
Sociodemographic RUCC 1
Loading
A priori Variable (Expected
Characteristic
Beneficial
Sign)
Loading
(Actual)
0.4689
Match
(Expected
versus
Observed)
Necessary To
Multiply Vector of
Loadings by-1? (Loading)A2
Yes 0.2199
Modified
Loadings
-0.4689
coefficient
E-l
-------
-0.3274
0.1072
0.3274
Percent unemployed
Harmful
Yes
0.1331
0.0177
-0.1331
Percent vacant housing
Harmful
Yes
Yes
0.0874
0.0076
-0.0874
Household income
Beneficial
Yes
-0.0141
0.0002
0.0141
Percent renter-occupied housing
Harmful
Yes
0.4463
0.1992
-0.4463
Percent creative class
Beneficial
Yes
-0.1604
0.0257
0.1604
GINI
Harmful
Yes
Count of occupants per room
Harmful
-0.1371
Yes
0.0188
0.1371
Violent crime
Harmful
-0.2386
Yes
0.0569
0.2386
Median household value
Beneficial
0.4002
Yes
0.1602
-0.4002
Percent Democratic
Beneficial
0.0929
Yes
-0.0929
Percent families less than
poverty level
Harmful
-0.4293
Yes
0.1843
0.4293
Percent bachelor's degree
Sociodemographic RUCC 2
Loading
A priori Variable (Expected Loading
Characteristic Sign) (Actual)
Beneficial 0.4621
Match
(Expected
versus
Observed)
Necessary To
Multiply Vector of
Loadings by -1?
Yes
Modified
Loadings
-0.4621
coefficient
Built (Overall)
Match
Loading (Expected Necessary To
A priori Variable (Expected Loading versus Multiply Vector of Modified
Characteristic
Sign)
(Actual)
Observed)
Loadings by -1?
(Loading)A2
Loadings
Vice-related environment
Harmful
+
0.2930
Yes
Yes
0.0858
-0.2930
Civic-related environment
Beneficial
-
0.3071
No
Yes
0.0943
-0.3071
Education-related environment
Beneficial
-
0.3495
No
Yes
0.1222
-0.3495
Health care-related environment
Beneficial
-
0.2798
No
Yes
0.0783
-0.2798
Negative food environment
Harmful
0.2280
Yes
Yes
0.0520
-0.2280
Positive food environment
Beneficial
-
0.3179
No
Yes
0.1011
-0.3179
Recreation environment
Beneficial
-
0.3590
No
Yes
0.1289
-0.3590
Social service-related environment
Beneficial
-
0.3629
No
Yes
0.1317
-0.3629
Traffic fatality rate
Harmful
+
-0.1751
No
Yes
0.0307
0.1751
Rate of low-rent + Section 8 housing
Harmful
+
0.0581
Yes
Yes
0.0034
-0.0581
Proportion of secondary roads
Harmful
-0.1777
No
Yes
0.0316
0.1777
Commute time
Harmful
+
-0.3329
No
Yes
0.1108
0.3329
Public transportation
Beneficial
-
0.0463
No
Yes
0.0021
-0.0463
Walkability score
Beneficial
-
0.1585
No
Yes
0.0251
-0.1585
Proportion green space
Beneficial
-
-0.0451
Yes
Yes
0.0020
0.0451
Built RUCC 1
Loading
A priori Variable (Expected Loading
Characteristic Sign) (Actual)
Vice-related Harmful " +" 0.2676
environment
Match
Necessary To
Multiply Vector of Modified
Loadings by -1? (Loading)A2 Loadings
Yes Yes 0.0716 -0.2676
(Expected
versus
Observed)
Civic-related
environment
Education-related
environment
Health care-related
environment
Beneficial
Beneficial
Beneficial
0.1238
0.2409
0.4189
No
No
No
Yes
Yes
Yes
0.0153
0.0580
0.1755
-0.1238
-0.2409
-0.4189
E-2
-------
0.3405
0.1159
-0.3405
Positive food
environment
Beneficial
Yes
0.3446
0.1187
-0.3446
Social service-related
environment
Beneficial
Yes
-0.1230
0.0151
0.1230
Rate of low-rent +
Section 8 housing
Harmful
Yes
0.0356
Commute time
Harmful
Yes
0.3516
0.1236
-0.3516
Walkability score
Beneficial
Yes
0.2057
0.2057
-0.2057
Civic-related environment
Beneficial
Yes
0.3856
0.3856
-0.3856
Health care-related environment
Beneficial
Yes
0.2752
0.2752
-0.2752
Positive food environment
Beneficial
Yes
0.3503
0.3503
-0.3503
Social service-related environment
Beneficial
Yes
0.0459
0.0459
-0.0459
Rate of low-rent + Section 8 housing
Harmful
Yes
Yes
Commute time
Harmful
Yes
0.3310
0.3310
-0.3310
Walkability score
Beneficial
Yes
Public transportation
Beneficial
0.2253
Yes
0.0508
-0.2253
Traffic fatality rate
Harmful
-0.2340
Yes
-0.2340
0.2340
Recreation environment
Beneficial
0.3484
Yes
0.3484
-0.3484
Proportion of secondary roads
Harmful
-0.1319
Yes
-0.1319
0.1319
Negative food environment
Harmful
0.2707
Yes
Yes
0.2707
-0.2707
Proportion green space
Beneficial
0.0253
Yes
0.0253
-0.0253
Public transportation
Beneficial
0.1111
Yes
0.1111
-0.1111
Education-related environment
Beneficial
0.2626
Yes
0.2626
-0.2626
Traffic fatality rate
Harmful
0.1978
Yes
Yes
0.0391
-0.1978
Proportion of
secondary roads
Harmful
0.0950
Yes
Yes
0.0090
-0.0950
Proportion green
space
Beneficial
-0.1065
Yes
0.0113
0.1065
Recreation
environment
Beneficial
0.2354
Yes
0.0554
-0.2354
Vice-related environment
Built RUCC 2
A priori
Variable
Characteristic
Harmful
Loading
(Expected Loading
Sign)
Match
(Expected
versus
(Actual) Observed)
0.0331
Yes
Necessary To
Multiply Vector
of Loadings
by -1?
Yes
(Loading)A2 Modified Loadings
0.0331
-0.0331
Negative food
environment
Built RUCC 1
Loading
A priori Variable (Expected
Characteristic
Harmful
Sign)
Loading
(Actual)
0.3239
Match
(Expected
versus
Observed)
Yes
Necessary To
Multiply Vector of
Loadings by -1?
Yes
Modified
Loadings
-0.3239
E-3
-------
0.1890
0.0357
-0.1890
Civic-related environment
Beneficial
Yes
0.3179
0.1011
-0.3179
Health care-related environment
Beneficial
Yes
0.2660
0.0707
-0.2660
Positive food environment
Beneficial
Yes
0.3644
0.1328
-0.3644
Social service-related environment
Beneficial
Yes
0.0697
0.0049
-0.0697
Rate of low-rent + Section 8 housing
Harmful
Yes
-0.3230
0.1043
0.3230
Commute time
Harmful
0.3542
0.1255
-0.3542
Walkability score
Beneficial
0.3102
0.0962
-0.3102
Civic-related environment
Beneficial
Yes
0.2742
0.0752
-0.2742
Health care-related environment
Beneficial
Yes
0.2524
0.0637
-0.2524
Positive food environment
Beneficial
Yes
0.2793
0.0780
-0.2793
Social service-related environment
Beneficial
Yes
-0.0178
0.0003
0.0178
Rate of low-rent + Section
housing
Harmful
Yes
-0.3546
0.1257
0.3546
Commute time
Harmful
Yes
0.3787
0.1434
-0.3787
Walkability score
Beneficial
Yes
Recreation environment
Beneficial
0.3212
Yes
0.1032
-0.3212
Traffic fatality rate
Harmful
-0.2197
Yes
0.0483
0.2197
Traffic fatality rate
Harmful
-0.2312
Yes
0.0535
0.2312
Recreation environment
Beneficial
0.3222
Yes
0.1038
-0.3222
Education-related environment
Beneficial
0.3278
Yes
0.1074
-0.3278
Proportion of secondary roads
Harmful
-0.1761
0.0310
0.1761
Public transportation
Beneficial
0.0777
0.0060
-0.0777
Proportion green space
Beneficial
-0.0418
Yes
0.0017
0.0418
Negative food environment
Harmful
0.2306
Yes
Yes
0.0532
-0.2306
Proportion of secondary roads
Harmful
-0.2054
Yes
0.0422
0.2054
Public transportation
Beneficial
0.0256
Yes
0.0007
-0.0256
Education-related environment
Beneficial
0.3285
Yes
0.1079
-0.3285
Proportion green space
Beneficial
-0.1370
Yes
Yes
0.0188
0.1370
Negative food environment
Harmful
0.1527
Yes
Yes
0.0233
-0.1527
Vice-related environment
Built RUCC 4
A priori
Variable
Characteristic
Harmful
Loading
(Expected Loading
Sign)
Match
(Expected
versus
(Actual) Observed)
0.2595
Yes
Necessary To
Multiply Vector of
Loadings by -1?
Yes
Modified Loadings
-0.2595
Vice-related environment
Built RUCC 3
A priori
Variable
Characteristic
Harmful
Loading
(Expected Loading
Sign)
Match
(Expected
versus
(Actual) Observed)
0.2724
Yes
Necessary To
Multiply Vector of
Loadings by-1? (Loading)A2 Modified Loadings
Yes 0.0742 -0.2724
Vice-related environment Harmful " +" 0.2595 Yes Yes 0.0673 -0.2595
Civic-related environment Beneficial 0.3102 No Yes 0.0962 -0.3102
Education-related environment Beneficial 0.3285 No Yes 0.1079 -0.3285
Health care-related environment Beneficial 0.2742 No Yes 0.0752 -0.2742
Negative food environment Harmful " + " 0.1527 Yes Yes 0.0233 -0.1527
Positive food environment Beneficial " -" 0.2524 No Yes 0.0637 -0.2524
Recreation environment Beneficial 0.3222 No Yes 0.1038 -0.3222
Social service-related environment Beneficial " -" 0.2793 No Yes 0.0780 -0.2793
Traffic fatality rate Harmful " + " -0.2312 No Yes 0.0535 0.2312
Rate of low-rent + Section 8 Harmful " + " -0.0178 No Yes 0.0003 0.0178
housing
Proportion of secondary roads Harmful " + " -0.2054 No Yes 0.0422 0.2054
Commute time Harmful " + " -0.3546 No Yes 0.1257 0.3546
Public transportation Beneficial " -" 0.0256 No Yes 0.0007 -0.0256
Walkability score Beneficial 0.3787 No Yes 0.1434 -0.3787
Proportion green space Beneficial -0.1370 Yes Yes 0.0188 0.1370
E-4
-------
Appendix Vh County Maps of Environmental
Quality Index 2006- 2010
Overall Environmental Quality Index by County 2006-2010
Percentile
¦ O-SOi
¦ 5th -20th
f"B~ 20,h - 40th
¦ 40th - 60th
I I 60«i - 80"'
80th - 95th
¦ 95lh - 100th
Air Domain Index by County 2006-2010
Percentile
* For orientation to the maps, low index scores (EQI and domain-specific) indicate higher environmental quality, and higher index
scores (EQI and domain-specific) mean lower environmental quality
I- ]
-------
Water Domain Index by County 2006-2010
Percentile
Land Domain Index by County 2006-2010
Percentile
¦ 0-5*
¦ 5th - 20th
¦ 20th ¦ 40th
E3 40,h - 60th
I] 60th - 80th
na 80th - 95th
¦ 95th- 100th
* For orientation to the maps, low index scores (EQI and domain-specific) indicate higher environmental quality, and higher index
scores (EQI and domain-specific) mean lower environmental quality
F-2
-------
Sociodemographic Domain Index by County 2006-2010
Percentile
Built Domain Index by County 2006-2010
Percentile
* For orientation to the maps, low index scores (EQI and domain-specific) indicate higher environmental quality, and higher index
scores (EQI and domain-specific) mean lower environmental quality
F-3
-------
Overall Environmental Quality Index Stratified by Rural-Urban Continuum Codes by County 2006-2010
! 0 - 5th Percentile
1 I I ~ ~ 5* - 20* Percentile
HI Ul Hi H 20th - 40th Percentile
H MEI |^| H 40th - 60th Percentile
¦I H |60th - 80th Percentile
80th - 95th Percentile
H 95th- 100th Percentile
RUCC1 = Metropolitan urbanized
RUCC2 = Non-metro urbanized
RUCC3 = Less urbanized
RUCC4 = Thinly populated
Built Domain Index Stratified by Rural-Urban Continuum Codes by County 2006-2010
I I I I I I 1 1 0 - 5th Percentile
H I 1 II 1 I1 I 5th - 20th Percentile
_J 20th - 40th Percentile
3] 40th - 60th Percentile
H |^| H 60th - 80th Percentile
H H 80th - 95th Percentile
H H |^| Hi 95th- 100th Percentile
RUCC1 = Metropolitan urbanized
RUCC2 = Non-metro urbanized
RUCC3 = Less urbanized
RUCC4 = Thinly populated
* For orientation to the maps, low index scores (EQI and domain-specific) indicate higher environmental quality, and higher index
scores (EQI and domain-specific) mean lower environmental quality
F-4
-------
Water Domain Index Stratified by Rural-Urban Continuum Codes by County 2006-2010
I 1 I 1 I 1 I I 0 - 5th Percentile
5th. 20th Percentile
I j Hi I I 20th - 40th Percentile
IH H HI 40th - 60th Percentile
Hi H 60th - 80th Percentile
|^| H Hi Hi 80th - 95th Percentile
H 95th- 100th Percentile
RUCC1 = Metropolitan urbanized
RUCC2 = Non-metro urbanized
RUCC3 = Less urbanized
RUCC4 = Thinly populated
Land Domain Index Stratified by Rural-Uiban Continuum Codes by County 2006-2010
I 1 1 I 1 1 I I 0 - 5th Percentile
_i 5th - 20th Percentile
| 20th - 40th Percentile
| Hi 40th - 60th Percentile
H H H Hi 60th " 30th Percentile
H H H 80th - 95th Percentile
| H 95th- 100th Percentile
RUCC1 = Metropolitan urbanized
RUCC2 = Non-metro urbanized
RUCC3 = Less urbanized
RUCC4 = Thinly populated
* For orientation to the maps, low index scores (EQI and domain-specific) indicate higher environmental quality, and higher index
scores (EQI and domain-specific) mean lower environmental quality
F-5
-------
Sociodemographic Domain Index Stratified by Rural-Urban Continuum Codes by County 2006-2010
I 1 I I \ I II I 0 - 5th Percentile
I 1 I FT 1 5th - 20th Percentile
I 1 Hi B I 20th - 40th Percentile
1^1 II II Hi 40th - 60th Percentile
H HI Hi 60th - 80th Percentile
HI Hi 8Qth - 95th Percentile
|^| H 95th - 100th Percentile
RUCC1 = Metropolitan urbanized
RUCC2 = Non-metro urbanized
RUCC3 = Less urbanized
RUCC4 = Thinly populated
Built Domain Index Stratified by Rural-Urban Continuum Codes by County 2006-2010
I 1 I I I I | | 0 - 5th Percentile
n ~ ~ EH 5th - 20th Percentile
20th - 40th Percentile
I 40th - 60th Percentile
|H H H 60th - 80th Percentile
H 80th - 95th Percentile
95th - 100th Percentile
RUCC1 = Metropolitan urbanized
RUCC2 = Non-metro urbanized
RUCC3 = Less urbanized
RUCC4 = Thinly populated
* For orientation to the maps, low index scores (EQI and domain-specific) indicate higher environmental quality, and higher index
scores (EQI and domain-specific) mean lower environmental quality
F-6
-------
Appendix VII: Quality Assurance
The approved Center for Public Health and Environmental
Assessment, Public Health and Environmental Systems Division,
Quality Assurance Project Plan for this project is "Creating an
Overall Environmental Quality Index," with Document Control
Number IRP-NHEERL/HSD/EBB/DL/2008-01-QP-1-7. An
internal EPA review of this report was conducted in April 2019.
An external peer review was conducted in March 2020.
The data sources used to create the EQI and the criteria used
to select the data sources are mentioned in this report in the
Development of the EQI 2006-2010 section.
Information about uses of the EQI, as well as strengths and
limitations of the EQI, is located within the Discussion section of
the report.
G-l
-------
SEPA
United States
Environmental Protection
Agency
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA
PERMIT NO. G-35
Office of Research and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
Recycled/Recyclable Printed on paper that contains a minimum of
50% postconsurner fiber content processed chlorine free
------- |