-
extracted text (extracttext:extracted_text)
-
SOCIOECONOMIC EQUITY
OF PUBLIC PARK ACCESS IN SEATTLE, WASHINGTON:
A SPATIAL ANALYSIS COMPARING
GIS BASED MEASUREMENT TECHNIQUES
by
Emma Giboney
A Thesis
Submitted in partial fulfillment
of the requirements for the degree
Master of Environmental Studies
The Evergreen State College
December 2017
©2017 by Emma Giboney. All rights reserved.
This Thesis for the Master of Environmental Studies Degree
by
Emma Giboney
has been approved for
The Evergreen State College
by
________________________
Shangrila Joshi, Ph. D.
Member of the Faculty
________________________
Date
ABSTRACT
Socioeconomic Equity of Public Park Access in Seattle, Washington:
A Spatial Analysis Comparing GIS Based Measurement Techniques
Emma Giboney
This study examined the municipal public park system and residents of Seattle,
Washington to determine if traditionally disadvantaged socioeconomic groups related to
income, age, race, and population density had equitable spatial access to parks and if the
methods used to measure spatial access affected the outcome of the results. Spatial access
was analyzed using Geographic Information Systems (GIS) to measure distance to the
closest park, the amount of park area within a quarter mile, and service areas around
parks. These three measures were completed with a variety of variables representing
residents (block and block group census units) and distance measurement (Euclidean and
network metrics). Through the use of the Mann-Whitney U statistical test, it was found
that most of the disadvantaged groups examined had equal or significantly better access
than their privileged counterparts. Vacant and renter occupied housing (used as proxies
for income), non-white residents, and residents in population dense areas all had equal or
significantly equitable access to parks as measured by park proximity, park acreage, and
park service areas. However, households with residents under 18 or over 64 had
significantly inequitable spatial access to parks than other households in at least one of
the three measures. By using both block and block group census units, as well as both
Euclidean and network metrics to run the spatial analysis and statistical tests, it was
determined that method variable selection does not always affect the outcome of the
results in the same way. Census unit selection affected the outcome of statistical results
more than spatial results, while the opposite was true for metric selection. Overall most of
the groups examined in this study had equal or equitable spatial access to parks, and the
method variables used had an inconsistent effect on the significance of the results.
!
TABLE OF CONTENTS
List of Figures......................................................................................................................v
List of Tables......................................................................................................................vi
Acknowledgements............................................................................................................vii
Chapter 1: Introduction........................................................................................................1
Chapter 2: Literature Review and Background Information...............................................7
Benefits of Parks......................................................................................................7
Park Access and Socioeconomic Equity................................................................16
Study Area History and Demographics.................................................................39
Chapter 3: Methods............................................................................................................53
Data Acquisition and Preparation..........................................................................53
Spatial Analysis.....................................................................................................55
Statistical Analysis.................................................................................................63
Chapter 4: Results..............................................................................................................67
Park Access and Socioeconomic Equity Results...................................................67
Method Variable Analysis Results.........................................................................77
Chapter 5: Discussion........................................................................................................89
Park Access and Socioeconomic Equity Discussion.............................................89
Method Variable Analysis Discussion...................................................................97
Limitations and Recommendations......................................................................104
Chapter 6: Conclusion......................................................................................................108
Works Cited.....................................................................................................................112
Appendix..........................................................................................................................125
!
iv
!
LIST OF FIGURES
Figure 1: Regional map of Seattle area..............................................................................40
Figure 2: Historical map of Seattle parks .........................................................................43
Figure 3: Close-up of Seattle neighborhood showing size difference of blocks and block
groups..................................................................................................................57
Figure 4: Close-up of Seattle neighborhood showing block groups with population
weighted centroids..............................................................................................58
Figure 5: Close-up of Seattle neighborhood showing blocks with geographic
centroids..............................................................................................................58
Figure 6: Close-up of Seattle neighborhood showing example measurements from block
group centroids to the nearest park boundary or boundary point using network
and Euclidean metrics.........................................................................................59
Figure 7: Close-up of Seattle neighborhood showing an example of the area within a
quarter mile of a block group centroid measured using a Euclidean and network
service areas. Demonstrates the size difference, and the effect on included park
area. ....................................................................................................................60
Figure 8: Close-up of Seattle neighborhood showing park service areas using Euclidean
and network service areas. Demonstrates the size difference and effect on total
units included inside area....................................................................................62
Figure 9: Seattle with the results from park proximity spatial analysis using blocks and
network distance. Results are split into quartiles, with dark color indicating
closer access........................................................................................................74
Figure 10: Seattle with the results from the amount of park area spatial analysis using
blocks and network distance. Results are split into quartiles, with dark color
indicating more park acreage within a quarter mile............................................75
Figure 11: Seattle with the results from park service area spatial analysis using blocks and
network distance. Results are split into blocks inside and outside the quarter
mile threshold......................................................................................................76
v
!
!
LIST OF TABLES
Table 1: Demographics of Seattle Washington from the 2010 Decennial Census............46
Table 2: List of spatial analyses performed using combinations of census units, metrics,
and access measures............................................................................................55
Table 3: List of all variables used in study........................................................................65
Table 4: Results of equity analysis using network distance and block units.....................72
Table 5: Results of equity analysis using network distance and block units.....................73
Table 6: Results of equity analyses using all combinations of census units, metrics, and
access measures..................................................................................................85
Table 7: Results of equity analyses grouped by spatial analysis and socioeconomic
variable to show consistency of results...............................................................86
Table 8: Descriptive statistic results of spatial analyses using all combinations of census
units, metrics, and access measures....................................................................87
Table 9: Results of method comparison analyses. Compares blocks with block groups,
and Euclidean with network metrics to determine if method selection plays a
significant role in spatial analyses......................................................................88
Table 10: The number and percentage of census units inside and outside of park service
areas when using Euclidean and network metrics..............................................88
!
vi
!
ACKNOWLEDGEMENTS
I would like to thank the staff and faculty of the Master of Environmental
Studies Program, especially my thesis reader Shangrila Joshi, for their
continued support throughout the thesis process. I would also like to thank
my friends and family, especially my children, who encouraged me to
complete my degree.!
!
!
vii !
CHAPTER 1: INTRODUCTION
Every year the proportion of the global population living in cities grows at a rapid
rate. Worldwide, more than fifty percent of people now reside in urban areas. In the
United States the statistic is even more dramatic at over eighty percent (United Nations,
2014). As urbanization continues, not only do urban areas expand, they become denser as
well (Kuittinen, Moinel, & Adalgeirsdottir, 2016). With urban space at a premium, open
and green spaces within cities are at risk of development (Kabisch, Qureshi, & Haase,
2015).
However, in the last few decades, as population has increased so too has the body
of research conducted on the benefits related to green spaces in cities. These benefits can
help negate many of the detriments of city life at the individual and community level
through improved health and environmental services among others (Kabisch et al., 2015;
Konijnendijk, Annerstedt, Nielsen, & Maruthaveeran, 2013; Sherer, 2003; Maller et al.,
2008). Yet studies from around the world indicate that due to the uneven spatial
arrangement of parks within a city, not all residents have equal access to parks or the
benefits derived from them. Socioeconomic factors such as race, income, age, and
religion have been shown to be the identifying characteristics that divide populations with
good and poor access to green spaces (Kabisch & Haase, 2014; Talen, 2010; Boone,
Buckley, Grove, & Sister, 2009; Cutts, Darby, Boone, & Brewis, 2009; Comber,
Brunsdon, & Green, 2008; Oh & Jeong, 2007; Wolch, Wilson, & Fehrenbach, 2005;
Nicholls & Shafer, 2001; Talen, 1997; among others).
Accessibility is a complex concept involving both spatial and social aspects.
However, since the spatial component of accessibility is more easily quantified, it is
1
!
commonly used as a way to measure access even though it does not take into account the
social aspects (Talen, 1997; Nicholls & Shafer, 2001; for more on social aspects of park
access, see Wang, Brown, & Liu, 2015). Although social aspects bring diverse and
important complexities to the study of access, the focus of this text will be solely on the
spatial aspects of access. Within this text, the terms ‘accessibility’, ‘park access’, and
other such phrases refer only to quantifiable measurements related to spatial proximity
and distribution unless specifically noted. These measures include, for example,
calculating the distance between residents and the nearest park or the amount of park area
within a certain distance from a resident. These spatial measurements are used as a
means to determine the equality and equitability of access. Equal access in terms of parks
is when different groups of people have the same access to parks, such as live the same
distance from, or live near the same amount of park area. In other words, there is an
insignificant difference between the levels of spatial access each group has. Equity, on
the other hand is typically used in this field as a way to refer to traditionally
disadvantaged groups having increased spatial access to parks in order to provide park
use opportunities that might not otherwise be possible due to social aspects of access
(Nicholls, 2001; Wang et al., 2015).
The methods of measuring access to public parks have continued to expand and
develop as the subject has become more studied and technology has improved. Using
geographic information system (GIS) computer software enables precise spatial analysis
through the measurement of spatial components such as distance and area. GIS can also
be used to run spatial analyses using demographic or other data associated with spatial
locations. Methods for measuring equitable access to public facilities has become more
2
precise and realistic as the body of literature has expanded and tested methods. However,
due to the relative newness of the discipline and the continual advancement of
technology, there is no set standard for how to measure and analyze park access.
Although many methods use the same protocols, or overlap in some way, the small
details of the methods, such as how people and parks are represented, or how the distance
is calculated, can sometimes effect whether results show inequitable, equal, or equitable
access (Higgs, Fry, & Langford, 2012).
In order to develop further understanding of the socioeconomic equity of park
access in an urban setting and how it accurately measured, further case studies are
necessary. Seattle, Washington is the largest city in the northwestern United States. It
has a well-established parks and recreation department as well as an extensive park
distribution. The population demographics have been shifting in recent years,
highlighting socioeconomic factors that could be at risk of inequitable access (United
States Census Bureau (USCB), 2010a; Seattle Parks and Recreation (SPR), 2014).
Seattle is a strong candidate for demonstrating the benefits of green spaces,
assessing the accessibility and equity of city parks, as well as testing the quality and
accuracy of GIS methods. A recent report from the city government provides information
on how Seattle parks benefit the residents and also which communities use the parks
more often. This data is helpful for understanding how access translates to use, and
enables the comparison of this access study to actual visitation data (SPR, 2014). Seattle
is known as ‘the emerald city’ because of its green landscape, and has a history of
strongly supporting parks (SPR, 2014). This case study provides insight on how evenly
3
!
the large park system is distributed among its communities and how spatial access may
be affecting potential park use.
Building upon previous research methods, I completed this park accessibility case
study in Seattle by combining several socioeconomic variables and accessibility
measuring methods in order to not only contribute new empirical data to the growing
body of research, but also provide new insight on which methods are most successful and
appropriate for measuring access. Specifically, a spatial analysis was completed with GIS
software using both Euclidean and network distance metrics, as well as census data
relating to race/ethnicity, housing occupation status, age, and population density at both
block and block group aggregate levels. The overall aim of this study was to determine if
traditionally disadvantaged residents of Seattle, Washington, regardless of location or
socioeconomic characteristics, live within an equitable distance to similar amounts of
public park acreage and whether the methods used to determine this affect the outcome of
the results.
This thesis is broken into several chapters. First I present a literature review
providing background information and relevant literature. This chapter contains three
large sections. The first pertains to the benefits of municipal parks, which fall into four
categories: health, social, economic, and environmental. This is followed by an in-depth
look at the theory and methods of measuring access to green spaces. My aim in this
section is to provide the information necessary for understanding the reasoning behind
the selection of methods and variables compared in this study. The last section of the
literature review provides history and other information about the study area.
4
Next the methods chapter lays out the specific methods and variables I used to
perform spatial and statistical analyses. This study implemented three measures of access:
nearest park, amount of park area, and park service areas. Each of these measures was
completed four times, using a combination of census unit (block and block group) and
metric variables (Euclidean and network distance). I completed an equity analysis by
combining the spatial results with socioeconomic variables to run a series of MannWhitney U statistical tests. Six variables were identified from previous research and were
used in this study to represent traditionally disadvantaged groups who benefit from park
access. These variables were (i) percent vacant housing units, (ii) percent renter occupied
housing units, (iii) percent non-white residents, (iv) percent households with residents
under 18, (v) percent households with residents over 64, and (vi) population density.
Next I offer the results and discussion chapters. Both of these chapters are divided
into two sections. The first section is a straightforward presentation of the spatial and
equity analysis using blocks and network distance variables. These variables were chosen
for the statistical equity analysis as the best representation of residents and how they
travel to parks. This selection is based on previous research performed by Nicholls
(2001), Higgs, Fry, and Langford (2012), and Talen (1997). The results of this analysis
show that over half of the socioeconomically disadvantaged groups examined have
statistically significant better spatial access as measured by park proximity, park acreage,
or park service areas than their traditionally privileged counterparts. In other words, these
groups were considered to have equitable access to parks. Unfortunately, the groups
related to age, those households with individuals under 18 and over 64 were found to
have inequitable access (significantly poorer access) in at least one of the measures.
5
!
The second section of the results and discussion chapters are related to an indepth method variable analysis. In this analysis I compared and contrasted the four
method variables (block and block group census units, and Euclidean and network
metrics) through descriptive statistics and statistical equity analyses. This analysis
produced mixed results, indicating that metric choice significantly affects the spatial
analysis and census unit choice does not. However the results of the series of equity
analyses show a pattern of census unit playing a larger role than metric choice in the
significance of results.
Lastly in the conclusion I provide a summary of the thesis to demonstrate the
contribution to the literature. Additionally the chapter includes the challenges I faced
during the research process and recommendations for future studies. This chapter
demonstrates how the research aims were met, and concludes that most of Seattle’s
traditionally disadvantaged groups have equal or equitable spatial access to parks, and
that although methodology can affect the spatial measurement, it has a limited and
irregular affect on determining if the disadvantaged groups have equitable access.
6
CHAPTER 2: LITERATURE REVIEW AND
BACKGROUND INFORMATION
BENEFITS OF PARKS
The benefits of green spaces in cities have become a widely researched topic. In
the past two decades, research articles from around the world have increased at a quick
rate. Many studies agree that the creation and enhancement of urban green spaces could
help ease the effects of urbanization and improve quality of life. The benefits found and
discussed in the vast amount of studies fall into four categories: health, social, economic,
and environmental (Chiesura, 2004; De Ridder et al., 2004; Kabisch et al., 2015; Maller
et al., 2008; Wolch, Byrne, & Newell, 2014). In order to demonstrate the importance of
park accessibility, this section of the text aims to highlight the relationship between the
benefits of parks and park accessibility. The health, social, economic, and environmental
benefits derived from parks and how they relate to proximity and accessibility will be
discussed.
Health Benefits
The health benefits from parks improve both physical and mental health. In a
study conducted by the Center for Disease Control (CDC), physical inactivity and obesity
were identified as two major epidemics plaguing American health. The report concluded
with a strong recommendation that access to locations for physical activity be enhanced
in order to help mitigate these specific concerns by promoting physical activity. It was
determined that having access to places where one could engage in physical activities
7
!
increased the likelihood of a person exercising three or more days a week by a full
twenty-five percent (Centers for Disease Control and Prevention, 2001). In an extensive
systematic review of park benefits compiled by Konijnendijk, Annerstedt, Nielsen, and
Maruthaveeran (2013), it was found that there is strong evidence to support the claim that
park access is linked with an increase in physical activity. Due to this increase in physical
activity, there is also strong to moderate evidence that park access reduces obesity.
Additionally, some evidence showed a reduced risk of stroke mortality and
cardiovascular and respiratory morbidity. Similar to the CDC, Konijnendijk et al.
conclude that there is sufficient evidence in the literature to establish that parks improve
health through increased physical activity (Konijnendijk et al., 2013). Parks improve
physical health by promoting physical activity. Increasing activity helps mitigate the
sedentary lifestyle and obesity issues in America. With the creation and enhancement of
parks, more urban residents would have access to these benefits.
Along with improving physical health, studies show that urban parks improve the mental
health of residents. Many studies have found that visiting parks reduces stress (Ulrich,
1981; Hartig, Mang, & Evans, 1991; Maller et al., 2008). In fact, one study determined
that the greater the amount of time spent in a park, the lesser the amount of stress
reported by residents (Hull & Michael, 1995). Additionally, studies have found that visits
to parks reduce anxiety, feelings of sadness, improves mood (Bedimo-Rung, Mowen, &
Cohen, 2005) as well as provide a sense of peace and tranquility (Kaplan, 1985). A few
studies found that access to parks or green spaces alleviates symptoms related to attention
deficit hyperactivity disorder in children (Taylor & Kuo, 2009; Markevych et al., 2014).
In a review of literature pertaining to how parks affect health, Bedimo-Rung et al. discuss
8
that simply knowing that a park is available as an escape from the urban setting can
benefit mental health (Bedimo-Rung et al., 2005). In another study, ten years of hospital
data was analyzed and determined that patients who had views of green space from their
rooms recovered more quickly from surgeries and required less medication than those
patients with views of buildings (Ulrich, 1984). According to the World Health
Organization, urbanization is associated with an increase in mental disorders. Since
visiting or having views of green spaces has been shown to improve mental health, it is
critical to create or maintain access to parks for vulnerable groups as population
increases.
Social Benefits
In addition to physical and mental health benefits, having access to parks can
provide social benefits to individuals and communities. Critically important, yet difficult
to quantify, the social benefits of parks are those that contribute to social interaction and
community cohesion or pride. In a study discussing the contribution of parks to
neighborhood social ties, Kaźmierczak (2013) emphasizes how parks facilitate
opportunities for social interaction and development of new relationships. By providing
space for visitors, parks enable a range of social interactions between users and can
promote a sense of community over time (Kaźmierczak, 2013). Related to encouraging
interactions, many studies mention the ability of parks to reduce intolerance of other
social groups. Increased globalization, urbanization, and migration have increased the
diversity of once socially homogenous neighborhoods or larger regions. Through shared
public space residents are exposed to people who have different backgrounds or ways of
9
!
life. Research indicates that experiencing social and ethnic diversity in public spaces
reduces intolerance and creates a sense of community cohesion (Konijnendijk et al.,
2013; Kaźmierczak, 2013). Some socioeconomic groups more than others benefit socially
from having access to parks. Age is an important characteristic of those who gain the
most benefit. Studies show that green spaces promote social integration among elderly
urban residents (Bedimo-Rung et al., 2005; Kaźmierczak, 2013). Additionally, young
children gain social benefits from urban parks through learning how to interact and
cooperate with peers. At-risk youths benefit from having a space to interact and spend
time with peers in positive ways (Sherer, 2003). Other groups that benefit socially from
green spaces include women and those groups with low income, poor health, limited
social opportunities, limited mobility, and high unemployment (Kaźmierczak, 2013).
Overall, studies show that parks promote both social interaction and community
involvement. However, many studies emphasize the lack of data in this area of research
and admit that more repetition is needed to support the claim that park access has social
benefits (Kaźmierczak, 2013; Konijnendijk et al., 2013).
Economic Benefits
Parks also provide economic benefits to individuals and communities. There are
many indirect economic benefits related to saving money due to other benefits mentioned
in this text, however the most researched and discussed direct benefit is the positive
impact parks have on property values. In a 2005 meta-analysis, Crompton states “people
frequently are willing to pay a larger amount of money for a home located close to a park,
than they are for a comparable home” (Crompton, 2005, p. 203). From this increased
10
monetary value, there is a subsequent increase in property tax paid to the city, which
benefits the larger community. Studies show that although in general park proximity has
a positive affect on property value, the optimal distance and orientation varied depending
on the park and neighborhood characteristics (Crompton, 2005; Nicholls & Crompton,
2005). However, it must be mentioned that studies find that not all residents view an
increase in property value as a benefit. Wolch, Byrne, and Newell (2014) describe how
an increase in property value can negatively affect residents in what they call the urban
green space paradox. In simple terms, when a new park is created in a park-impoverished
area, the “housing cost escalation can potentially lead to…the displacement and/or
exclusion of the very residents the green space was meant to benefit” (Wolch et al., 2014,
p. 235). In other words, because of the increase in taxes or rent costs, the residents may
be financially “forced to leave their communities, ending up in less desirable
neighborhoods with similar park-poverty problems”. According to the authors, the
challenge produced by this paradox is creating areas that are “just green enough” to give
benefits to the residents without causing gentrification (Wolch et al., 2014, p. 235).
Despite this viewpoint, most research sees increased property value as a benefit overall.
Environmental Benefits
Lastly, parks provide a wide variety of environmental benefits to individuals,
neighborhoods and entire cities. Most benefits are directly related to the lack of buildings,
ample vegetation and permeable ground of parks. Many studies have been conducted on
benefits such as improved air quality, temperature regulation, and rainwater mitigation. A
handful of other benefits are mentioned in the literature, including carbon sequestration
11
!
and noise reduction. The environmental benefits of urban parks have an intrinsic value in
addition to providing better health and economic savings to cities, communities, and
individuals.
One of the most heavily researched environmental benefits of urban green spaces
is the filtering of air pollution to improve air quality. Air pollution is caused by
urbanization, transportation, energy production, and industrialization, among others
(Cohen, Potchter, & Schnell, 2014; Gómez-Baggethun et al., 2013; Cohen et al., 2004).
The majority of these pollutants are small enough to be inhaled into the lungs (Cohen et
al., 2004), making air pollution a major concern not only for environmental quality but
also human health at the local and global level (Gómez-Baggethun et al., 2013;
Konijnendijk et al., 2013). However, green spaces mitigate air pollution in urban areas.
Vegetation improves air quality through the filtering of pollutants through plant
respiration as well as retaining pollutants through direct interception and deposition onto
the plants surface (Gómez-Baggethun et al., 2013; Norwak, 2002). Studies from around
the world show that concentrations of pollutants are lower inside of vegetated green
spaces than other urban areas (Cavanagh, Zawar-Reza, & Wilson, 2009; Yin et al., 2011).
The rate of pollution mitigation depends heavily on the characteristics of the vegetation
and green space (Bolund & Hunhammar, 1999). Most studies rely on modeling, however
a few projects use on-site data collection to demonstrate vegetative filtering abilities
(Konijnendijk et al., 2013; Cohen et al., 2014).
Parks can help to reduce atmospheric carbon in urban areas. Carbon in the
atmosphere is sequestered into plants through photosynthesis to make new plant biomass.
From plants it continues through the carbon cycle, being stored in oceans, land, and
12
rocks. Carbon can be released back to the atmosphere through anthropogenic causes
including burning fossil fuels, cement production, land use change, among others
(Chapin, Matson, & Mooney, 2002). In recent years, urban greenery as a means to
mitigate climate change has gained some popularity. There is some evidence that
vegetation in urban areas can absorb some atmospheric carbon in cities. Unfortunately, in
the vast majority of studies the total carbon sequestered in urban green spaces does not
negate the total emissions of carbon (Velasco, Roth, Norford, & Molina, 2016). However,
parks and green spaces in urban areas store carbon that could otherwise be in the
atmosphere contributing to climate change. Although parks do not sequester enough
carbon to negate the vast amount of emissions in urban areas, the amount of carbon parks
do store does have significant environmental and monetary value (Chapin, Matson, &
Mooney, 2002; Gómez-Baggethun et al., 2013; Nowak, Greenfield, Hoehn, & Lapoint,
2013).
Due to climate change, temperatures and microclimates are becoming less
predictable and more extreme (Gómez-Baggethun et al., 2013). This effect can be
exacerbated in urban areas, which tend to become islands of heat compared to
surrounding cooler countryside (Bowler, Buyung-Ali, Knight, & Pullin, 2010;
Konijnendijk et al., 2013; Chang, Li, & Chang, 2007). In addition to environmental
impacts, higher urban temperatures affect public health and the economy through an
increased risk of heat-related illness and higher energy consumption (Bedimo-Rung et al.,
2005; Bolund & Hunhammar, 1999). However, studies show that urban green spaces help
to regulate local temperature and microclimate in urban areas through various means
(Konijnendijk et al., 2013). Jim and Chen (2009) mention differences in solar radiation,
13
!
wind speed, humidity, and terrestrial re-radiation as influences on microclimate. Several
studies highlight plant evapotranspiration as a key mechanism in the cooler temperatures
of green spaces (Bowler et al., 2010; Jim and Chen, 2009; Sherer, 2003). Bowler et al.
(2010) point out that the regulatory mechanisms depend greatly on the characteristics of
the green space such as vegetation size and type, park size, among others. Although there
are varying results on the overall affect on temperature, research agrees that parks have
less extreme temperature variations than surrounding areas (Konijnendijk et al., 2013;
Cohen et al., 2014; Bowler et al., 2010; Chang et al., 2007).
Green spaces are beneficial to urban areas by reducing rainwater surface runoff.
Pavement and buildings are impervious surfaces that do not soak in water like vegetated
land does. This change in land cover affects the way in which water and surface pollution
collects and flows within and out of the city (Konijnendijk et al., 2013). The increase in
impermeable surfaces leads to an increased volume of surface water runoff, and in turn
increases the risk of flooding (Gómez-Baggethun et al., 2013). Green spaces can mitigate
some runoff and regulate drainage (Yang, Zhang, Li, & Wu, 2015). The rate at which
rainfall reaches the ground is controlled through interception by the leaf canopy (Yang et
al., 2015). Once rainfall reaches the ground, soil acts as a sponge to store and percolate
the water (Gómez-Baggethun et al., 2013). These mechanisms decrease flood hazards and
reduce stress on drainage infrastructure, which saves money at the individual and city
level (Gómez-Baggethun et al., 2013; Zhang, Xie, Zhang, & Zhang, 2012). In a study
conducted by Kaźmierczak and Cavan (2011), it was found that areas of Manchester,
England with a lower percentage of green space were more susceptible to surface water
flooding (Kaźmierczak & Cavan, 2011). In a 2015 study, it was found that green spaces
14
absorbed 88% of rainwater in Yixing City, China (Yang et al., 2015). Another study from
China found that the amount of money saved by the city of Beijing through the storage
and processing of water by green spaces was equivalent to 75% of the annual costs to
maintain the city’s green spaces (Zhang, et al., 2012). Although there is strong evidence
that green spaces play an important role in runoff reduction, most studies rely heavily on
models rather than experimental data. More empirical data is needed to strengthen the
evidence and gain insight into what characteristics of green spaces make the greatest
impacts (Konijnendijk et al., 2013).
One detriment of living in an urban area is higher levels of noise pollution.
Excessive noise levels can negatively affect wildlife as well as human mental and
physical health as well as behavior. High noise levels can cause stress, and in general is
considered to be one of the most detrimental factors to urban residents’ daily lives.
(González-Oreja, Bonache-Regidor, & de la Fuente-Díaz-Ordaz, 2010). Green spaces
help to attenuate noise pollution by varying degrees depending on space size and
vegetation amount. Vegetation and soil absorbs, deviates, reflects, and refracts sound
waves (González-Oreja et al., 2010). Studies using on site measurements show that green
spaces are quieter than nearby streets and paved open spaces (Cohen et al., 2014). Urban
parks are a space for urban residents to escape from noise. However, some research
advocates for policy change for as the best long-term solution for reducing urban noise
levels (González-Oreja et al., 2010).
As demonstrated by the literature I reviewed in this section, there is an extensive
amount of benefits provided by green spaces in cities. Research from around the world
15
!
shows that the benefits provided by parks has become a global interest. However, for
many of these benefits to be realized, accessibility or proximity to a park is a necessity.
While some benefits are gained by all, such as carbon sequestration, other benefits
require a nearby park, such as reduced noise pollution or increased property value. Some
benefits such as less variable temperatures and increased exercise can only be obtained
by visiting the park. Now that the benefits of parks have been laid out and it is understood
why park accessibility is a subject worthy of attention, I turn to a discussion of how
equitable access to parks is measured.
PARK ACCESS AND SOCIOECONOMIC EQUITY
As shown above, it has become well established in the last quarter century that
parks are beneficial to urban residents. A key part of gaining the majority of these
benefits is living near or having access to parks. However, parks may not be distributed
equitably to all residents. The spatial relationship between the location and characteristics
of both residents and parks is the underlying focus of the study of park accessibility. As
time passes and technology advances, more complicated questions about this relationship
can be asked and studied. However, there are many different ideas of what access is and
what it means to be equitably distributed. In this section, I provide an overview of park
accessibility literature to help in the understanding of spatial park access and how they
are measured. This section is broken into five subsections, each focused on a different
aspect of measuring the equity of park access. First, more in-depth definitions of access
and equity will be provided. In the second subsection, I introduce the data and technology
used to measure access and equity. Third, I discuss the variables that are used to represent
16
people and parks, as well as different types of metrics used to measure access. Example
studies are given to illustrate the range in variable use. In the fourth subsection, three
different GIS based measurement protocols used in spatial analyses are introduced, with
examples of previous studies using each. The last subsection focuses on how equity is
analyzed with a discussion of the results from previous studies. This overall aim of this
section is to provide a large amount of technical information as clearly as possible so as
to provide a clear picture of how equitable access is measured and analyzed.
Definitions of Access and Equity
Accessibility is a complex concept with many components that can be studied in
different ways. There are spatial and social aspects of access that can promote or hinder
access depending on the individual and their circumstances. Although the social aspects
of access are fascinating and well worth looking into, the majority of park access studies
within the field of urban geography focus purely on the quantifiable spatial aspects (for
more on social aspects of park access, see Wang, Brown, & Liu, 2015). For this text,
while acknowledging social aspects bring diverse and important complexities to the study
of access, the focus will remain on the spatial aspects and the term ‘access’, will refer to
spatial access, unless otherwise noted.
A second vital part of the study of park accessibility is looking at how spatial
access varies for residents and communities. This raises the concept of equity and what it
means in terms of access to parks. For access to public parks, two forms of equity are
most often used: equality and need-based equity. Equality is the equal opportunity to
access parks regardless of location or socioeconomic characteristics. Need-based equity
17
!
refers to residents or neighborhoods that are disadvantaged socioeconomically (typically
based on race/ethnicity, age, income, religion, and others) having increased spatial access
to parks in order to provide opportunities that might not otherwise be possible due to
social aspects of access. Many studies determine that access is equitable when a
disadvantaged group has better access than its privileged counterpart, and equal access
when the two groups have the same ease of access (Nicholls, 2001; Kabisch & Haase,
2014; Talen & Anselin, 1998; Comber et al., 2008). In this text, the term ‘equity’ will
follow these previous studies and refer to need-based equity based on disadvantaged
socioeconomic characteristics. Building upon these definitions of access and equity, a
‘spatial analysis’ measures the spatial relationship between residents and parks, while an
‘equity analysis’ uses statistics to analyze the differences in spatial accessibility by
socioeconomic group with the hope that all groups have equal access, or that traditionally
disadvantaged groups have better access.
Data and Technology
The measurement of spatial access to parks can be very complex, but underlying
all methods there are three important aspects: parks, people, and a spatial measurement.
Methods can be wide ranging and answer different questions about spatial accessibility,
but each one is based on a spatial analysis of the relationship between parks and people.
Therefore, it is important to discuss where data on people and parks come from and what
technologies make spatial analysis possible.
For most park access studies, data regarding people in the United States is
obtained from the U.S. Census Bureau. A wide range of data is available to the public,
18
including data on socioeconomic characteristics like race, age, gender, and housing
status, among others (Williams, 2011). These attributes become important for comparing
the equity of access between different locations or socioeconomic groups. However, this
data is not available at individual or household level. The smallest scale that any census
data is tabulated and publicly available is at the census block level, but some data is only
tabulated at block group or census tract levels. Census blocks are small areas of land
bounded by streets, railroads, streams, or other physical or political features (in urban
areas, they can often be, but are not always, single city blocks). One level above census
block is block group, which is an aggregate of several blocks, and an aggregate of block
groups makes up a census tract (USCB, 1994). Many studies use the block group level or
larger, depending on how large of a study area the analysis is covering or how much
detail the authors want. For example, Boone, Buckley, Grove, and Sister (2009) used
census data regarding race and ethnicity at the block group level to determine if one
category had better park access than the other. However, when determining if lowincome residents had equitable access to parks Boone et al. (2009) used data at the census
tract level since income data is not available at a smaller unit. The census collects a wide
range of data, and because it is associated with a physical location it can be used for
spatial analysis (Williams, 2011).
Other data used in a spatial analysis of park access is information on parks, as
well as information on the study area as a whole. This includes spatial data of where
parks are located as well as their sizes, and a variety of other attributes. In addition to
park data, a vast amount of data on road networks, blocks, city limits, and many other
spatial attributes are widely available online at a local level, especially for large urban
19
!
areas. Additionally, some studies include spatial data collected in the field, either because
it is an older study from before data was easily accessible online, or because the study
design requires more specific details that are not digitally available. For example,
Kaczynski, Potwarka, & Saelens (2008) collected data at thirty-three parks on what
facilities and amenities were present. Today however, most of the commonly used data is
available online and, using sophisticated computer programs, can be used in spatial
analyses.
Geographic information system (GIS) is a computer-based tool that enables the
storage, retrieval, manipulation, analysis, and graphic presentation of spatial data. It uses
complex databases and software to help the user understand spatial relationships,
patterns, and trends that would otherwise be overlooked in spreadsheet form. GIS uses a
coordinate system to spatially link data to a location. Similar to how a paper map could
have both roads and topography, GIS uses what are known as layers to display multiple
levels of geographically linked information. Each layer is a visual representation of a
database that can contain data from a wide variety of fields (e.g. urban planning, geology,
social services, political science etc.). Location is the common identifier that cuts through
every layer, making it possible to access attributes from several databases, based on a
location. It is possible to go the other direction as well, instead of accessing the attributes
of a single location, every location with a given attribute can be found. GIS is a complex
tool, but even using the most basic functions can enable an analysis between layers
making it possible to find profound spatial relationships (Leslie et al., 2007; Nicholls,
2001).
20
With the use of GIS, it is possible to digitally represent a city with a road network,
parks, and census units with census information, so that the relationship between parks
and census units, and therefore the resident characteristics, can be analyzed. With an
understanding of where the data concerning parks and people is retrieved from, and the
tools available for analysis, I turn to a discussion of how previous studies have
implemented this data in GIS to measure access.
Measuring Park Access: People, Park, and Metric Variables
Measurements of access range from simple to complex, and can answer different
questions about access. For example, one of the most basic and common measurements
of access is simply the distance from a person to the nearest park. More complicated
measurements include how much area of parks is within a certain distance from a person.
Another way of thinking about access is shifting the focus from the resident to the park,
for example by looking at the people within a set distance from the park. These
measurements of access could each answer a different question: how close is the nearest
park? How much park acreage is within walking distance? What community does this
park serve? With the help of GIS, these measurements are made possible on a large scale,
but it is still important to understand what the calculation is measuring and how to
interpret, or apply statistical analysis to the results. Because this thesis aims to not only
perform a spatial and equity analysis in Seattle, but also to complete a method
comparison analysis, it is important to discuss the variables and measures that are used to
determine spatial access. In this section I will focus on the variables used, whereas the
21
!
measurement protocols that implement these variables will be discussed in the next
section.
As mentioned previously, the basis for all methods are people, parks, and a spatial
measurement. With the knowledge of where data on people and parks is derived, and the
computer program available to make measurements, it is now important to discuss how
access can be measured, both in a theoretical and technical way. Higgs et al. (2012)
provide a starting point for thinking technically about access. The authors describe the
three fundamental elements of any method as:
(1) an origin point, representing the geographical location of the population
potentially seeking to access green space;
(2) a destination point, representing the geographical location of the green space;
and
(3) a distance measurement taken between these two points. (p. 328)
However the first complication when implementing a measurement of access in GIS is
how to precisely define or represent each of these aspects. Due to the nature of the data
and technology available, it is nearly impossible to measure spatial access with perfect
precision. Using the example from above, measuring from households to the nearest park,
ideally the measurement would be the distance from every household along the route the
residents would walk or drive to the point of the park that they potentially would access.
Herein lies a methodological limitation; it is not feasible to know or supply this
information to GIS on a wide scale. So one must choose proxies for the origin point,
destination point, and measurement to create a model of access as opposed to literal
22
measurements. The proxies that are selected have the potential to impact the measures of
access (Higgs et al., 2012; La Rosa, 2014).
Location of the population is typically estimated through the use of census areas.
Some studies that utilize respondent surveys are able to use the specific household
address of participants (Kaczynski, Potwarka, & Saelens, 2008; Koohsari, Kaczynski,
Giles-Corti, & Karakiewicz, 2013; Markevych et al., 2014; Wang et al., 2015), but most
accessibility studies do not use this method. It is more common to use geographically
referenced census units as the base for choosing a point of origin. Depending on the scale
of the study area or what level the desired census data is tabulated at, previous studies
have used census tracts (Boone et al., 2009; Higgs et al., 2012; La Rosa, 2014; Talen &
Anselin, 1998; Wolch et al., 2005), block groups (Abercrombie et al., 2008; Boone et al.,
2009; Cutts et al., 2009; Nicholls, 2001), and even blocks (Talen, 1997; Boone et al.,
2009). In addition to choosing a scale of census unit to use, GIS requires a point of origin
to perform a measurement, so no matter what the scale of census unit used to represent
the population, there must also be a single point within a census unit polygon to initiate
the start of a measurement. Most studies use geometric (Comber et al., 2008; La Rosa,
2014; Nicholls, 2001; Nicholls & Shafer, 2001; Talen, 1997; Talen & Anselin, 1998) or
population-weighted centroids (Boone et al., 2009; Higgs et al., 2012; Nutsford, Pearson,
& Kingham, 2013), depending on what data is available or a preference by the author.
Geometric centroid is a mathematical concept and describes the mean center position of
all the points in a polygon. In terms of census units, it is a single point created by
averaging the coordinates of the boundary of the polygon. When a centroid is weighted
by population, it uses the location of residences within the census unit to calculate the
23
!
mean position of the population within the unit (Bolstad, 2012). As Higgs et al. (2012)
notes, very little research has been done on what effect the selection of geometric or
population-weight centroids has on the analysis, but it possibly varies depending on what
scale of census area is being used.
Accurately representing the location of parks is perhaps even more complex than
population. The options most commonly used are geometric centroid (Kaczynski et al.,
2008; Kaczynski, Potwarka, Smale, & Havitz, 2009; La Rosa, 2014; Talen, 1997; Talen
& Anselin, 1998; Wang et al., 2015), park entrances (Kabisch & Haase, 2014), access
points (Comber et al., 2008; Nicholls, 2001), or park boundary (Boone et al., 2009;
Koohsari et al., 2013; Wolch et al., 2005). In this case, the geometric centroid is a single
point generated by a GIS tool using the coordinates of the park polygon to find the mean
center of the park (Bolstad, 2012). Some authors advise against the use of centroids since
the distance to the center of a large park would greatly skew the measurement (Boone et
al., 2009). Park entrances are the officially designated points of entry. Access points
include the official entrances, along with other point that the author identifies (Higgs et
al., 2012). For example, Nicholls visited each park in the study area to pinpoint where
access points were located (Nicholls, 2001). In addition to site visits, Higgs et al. (2012)
used survey maps and aerial photos to determine where access points were. Park
boundary is even more complicated depending on the type of measurement used. Some
measurements can simply use the polygon shape as the destination, whereas others
require an actual point to end the measurement. Koohsari et al. (2013) chose to distribute
points along the boundary at a set distance apart as a workaround methodology that do
not work with polygons. However, using the boundary comes with assumption that parks
24
are accessible from every angle, which might create some error, but perhaps not as much
as other options (Nicholls, 2001).
Lastly, there are options for how distance is measured. There are primarily two
metrics used in GIS park access studies: Euclidian (Boone et al., 2009; Cutts et al., 2009;
Kabisch & Hasse, 2014; Kaczynski et al., 2008; Kaczynski et al., 2009; Nutsford et al.,
2013; Wang et al., 2015) and network distance (Comber et al., 2008; Koohsari et al.,
2013; Talen, 1997; Talen, 1998; Talen & Anselin, 1998; Trust for Public Land (TPL),
2017). Euclidian distance is a straight line between two points. Euclidian is only an
approximation of the actual distance resident would need to travel to visit a park. It is ‘asthe-crow-flies’ and is not impeded by any barriers; whereas network distance measures
along a road, trail, or other specified network from one point to the other. It uses the
network to more-or-less follow the shortest route that a resident could take to a park
(Nicholls, 2001). For example, a resident might appear to live very close to a park when
measuring in Euclidean distance, but if there were a river, or railroad, or freeway between
the residence and the park, the network distance would potentially need to detour a great
distance in order to reach the same park. This is an extreme case; the true distance a
resident would travel to reach a park is probably in between the two measurements. Cutts,
Darby, Boone, and Brewis (2009) mention that residents can use informal paths and
shortcuts through neighborhoods that make the actual walking distance less than the
network distance. One benefit of Euclidian is that it is easy to understand and implement
in GIS, whereas network distance requires a lot more set up and input from the user (La
Rosa, 2014).
25
!
Many studies not only aim to gain insight on the accessibility of the parks, but
also on which methods are the most affective. Some evaluate previously established
methods, experiment with new strategies, or test methods against each other in order to
gain insight on what factors may play an influence. Many studies have the primary focus
of how choosing the variables discussed above affect the results, and use a case study
only to test the methods, more than for the results of the access study itself. Because this
study aims to both evaluate Seattle in an access and equity analysis as well as compare
the results found using different variables, it is important to note what previous studies
have found when comparing methods.
In 2001, Nicholls was able to compare Euclidean and network distances by
completing a park access and equity analysis. Using each metric, the author examined the
households that were within a half-mile of a park in the study area (the details of this
protocol will be discussed in the subsequent section, with the focus here being on the
variable selection). Interestingly, the author used the geographic centroid for the origin
point of Euclidean distance, but park access points for the network distance. Regardless
of this origin point difference, results show that when using Euclidean distance to
calculate the measurement, nearly a fifth more of the population was within the half-mile
distance. However, Nicholls then performed an equity analysis in which she compared
socioeconomic data from the households within and outside the half-mile distance from
parks and unexpectedly found that despite the metric having a large affect on the spatial
analysis, it only made a small difference in the significance of the equity analysis results.
Meaning that Euclidean and network metrics produced similar results as far as what
communities had better access than others (the detailed results of this analysis will be
26
discussed in a subsequent section, the aim here is to highlight that variable selection plays
a role in results) (Nicholls, 2001). Also in 2001, Nicholls worked with Shafer to complete
a study using the same methods and found similar results, finding that the metric used
had little affect on the statistical significance of the results of the equity study, but that a
higher percentage of the population was within a half-mile of a park (Nicholls & Shaffer,
2001). These studies demonstrate how the variables used to measure access can affect
some aspects of an analysis more than others.
In their study from 2012, Higgs et al. perform an extensive comparative analysis
of multiple methods. Not only are Euclidean and network distances compared, but also
the representation of green spaces as centroids, access points, as well as boundary points.
The aim of their study was to “evaluate the degree to which outcomes are influenced by
the details of the approach adopted” (p. 330). Using population-weighted centroids of
census output areas (the United Kingdom equivalent to block groups) as the origin point,
the authors performed six analyses by combining each green space representation with
each distance metric to find the distance to the closest park. Then by performing a
statistical analysis, the authors were able to determine if the measurements using each
variable were correlated. The authors’ most profound result was the variation in the green
spaces that were determined to be the closest depending on what destination target was
used (e.g. when using centroid, one green space was a shorter distance away and then
when using a boundary point a completely different green space was considered closer).
When using the Euclidean metric, Higgs et al. found that there were notable differences is
the distances depending on the choices of destination point. The measurements were
positively linked, but over twenty percent of the closest green spaces selected were
27
!
different depending on the destination used. Similar but more extreme results were found
when using network distance; The differences in distances were even greater when using
network distance with different destination targets, although they did have a positive link,
but less so than Euclidean. The green spaces being selected as closest were different over
forty-four percent of the time, much greater than when using Euclidean distance. In
comparing Euclidean and network distances to each other, the authors discovered that
there was a positive link between the measurements to each destination target variable,
but that the green spaces considered closest were different over half of the time (Higgs et
al., 2012).
These studies highlight the importance of proxy and metric selection. Depending
on the size of study area, the question being researched, and the amount of time available,
different methods could be considered appropriate. Higgs et al. emphasize that regardless
of which method is selected it is important to carefully report what metrics and proxies
are being used. Additionally, they recommend using a range of methods in order to
accurately measure accessibility (Higgs et al., 2012).
The aim of this section was to introduce the variables that come into play when
measuring access. How people and parks are represented, as well as what metric is used
can have an affect on the results of measuring the equity of access to parks (Higgs et al.,
2012). With an understanding of these variables in place, I will now discuss the specific
spatial analysis methods that these variables can be used in to measure access.
28
Measuring Park Access: Protocols for Spatial Analyses
There are a handful of spatial analysis methods that are commonly used to
measure park accessibility for equity analyses. The variables discussed above can be
implemented in GIS in many ways to answer different questions and aspects of access
and equity. Below I will discuss a few of the most popular spatial access measurement
protocols, giving examples of studies that used them in order to highlight how the same
methods can be implemented using different variables. These methods measure the
distance to the nearest park, the amount of park area within a certain distance, and the
service areas around parks. This section will remain focused on the GIS based
measurement techniques implemented in spatial analyses by previous studies, while the
next section will focus on how these results were then implemented in statistical analyses
to determine the equity of access.
A measure that is very commonly used, and offers a foundation for many other
access and equity analyses is finding the distance to the nearest park. This method can be
completed using any combination of the variables described above, but the aim of
measuring proximity remains the same. Studies often use this method to determine if
there is park space within walking distance, which has become standardly measured at
400 meters or one-quarter mile (Boone et al., 2009; Cutts et al., 2009; Harnik & Simms,
2004; Higgs et al., 2012; Nicholls, 2001). Finding the nearest park can provide the base
for many other analyses; with equity analyses being just one type this measure of access
can be applied too. For example, Nutsford, Pearson, and Kingham (2013) performed a
study in Auckland, New Zealand, examining the relationship between park proximity and
mental health disorders. Using a road network, the distance from the population-weighted
29
!
centroid of each meshblock (similar to block groups in the United States) to the closest
green space was calculated. Unfortunately, as is not uncommon, how the park was
represented was not defined in the publication. Regardless, Nutsford et al. found that on
average it was 210 meters to the nearest usable green space, falling well within what is
commonly considered a walkable range. Additionally when coupled with anxiety/mood
disorder treatment data, it was found that there was an association with distance to the
nearest usable green space (Nutsford et al., 2013). However, knowing the distance to the
nearest park is a limited measure of spatial access. A small neighborhood park is treated
equal to a large regional park. But this is still an important access measure since, as both
Wolch et al. (2005) and Boone et al. (2009) mention, it is better to have walkable access
to any park than no park at all.
Building upon measuring to the closest park, calculating the amount of park area
within a specified distance from residents is another measure of spatial park access. This
makes it possible to compare how much park area different socioeconomic communities
have. This measurement goes beyond asking if the distance to a park is equitable,
meaning disadvantaged groups have closer proximity to their nearest park, but also if the
amount of park area is equitable, meaning the amount of nearby park area is greater than
that of privileged groups. The amount of accessible park area is calculated in GIS by
creating service areas using Euclidean or network distance. Regardless of the metric used,
both use an origin point and measure outward in all directions to a specified distance in
order to create a polygon around the origin point representing the area that is accessible
within the specified distance (Bolstad, 2012). For this access measure, the center of the
service area is the point representing the residents, typically the geographic or
30
population-weight centroid of a census area (as discussed in the previous section), or in
some cases a point representing the home of a survey participant. To determine what park
area is considered inside of the service area, different criteria such as using a park
centroid or park boundary can be used, and then the entirety of the area is typically
included, even if the majority of the park polygon is outside of the park service area. An
example of a study using park area to measure access is Kaczynski, Potwarka, Smale, and
Havitz (2009). The aim of this study was to determine if there was a relationship between
proximity to park area and park-based physical activity. Due to part of the study
involving surveys, the authors were able to use respondents’ addresses as starting points
for the analysis. The authors used service areas of one Euclidean kilometer (0.62 miles)
and if a park’s geometric centroid was within the bounds, the entire area was included
(Kaczynski et al., 2009). In contrast to Kaczynski et al. (2009) use of the Euclidean
metric, Talen (1997) chose to use network distance to measure the amount of park area
within both one and two miles of census blocks in Pueblo, Colorado, and Macon,
Georgia. Similar to Kaczynski et al. (2009), Talen (1997) used the geometric centroid of
parks being within the boundary as the criteria for inclusion of the entire park area.
Boone et al. (2009) also used park area as a measure of access, but used a much smaller
service area of only one-quarter Euclidean mile to represent the amount of park area
within walking distance. Frustratingly, the authors did not specify the criteria used to
determine if a park’s area was included (Boone et al., 2009). Each of these studies
demonstrates how measuring park area in an spatial analysis can use the same access
measure but can be tailored through variable selection to suit the needs of the author.
31
!
While these studies used a variety of metrics and distance thresholds, they all measured
the same aspect of access, park area.
A third measure of access uses GIS based measurement techniques to spatially
analyze the populations living inside and outside a certain distance threshold of parks in a
study area. This is achieved by placing service areas, as described in the previous
measure, around each park using a specified metric. Because parks vary in size so
greatly, using a centroid is not common, due to the result that the park area would be
included within the service area. Due to the technical requirements and specific
calculations processed, GIS can easily create service areas using the boundary as the
starting point using the Euclidean metric but cannot when using the network metric,
which must start from a single point. However, Nicholls (2001) used a creative method to
construct network service areas that would closely match a true boundary starting point.
The author created half-mile network service areas for each access point of a park and
then joined them together to create a single area for the entire park (Nicholls, 2001).
Service areas could be created using the network metric in this manner using boundary
points or entrances as well. An additional example of a case study using network park
service areas but using different variables is Kabisch and Hasse (2014). These authors
created 1500-meter service areas using park entrances as the origin points in their
analysis of a large park in Berlin, Germany. Unlike Nicholls (2001) and Kabisch and
Hasse (2014), Boone et al. (2009) chose to use the Euclidean metric to create quartermile service areas around the boundaries of park in Baltimore, Maryland. Each of these
studies created park service areas but used different origin points, distance thresholds, or
32
metrics. Once again, this measure, just like the previous two, demonstrates how the same
GIS protocol can be used with different variables to best fit the aim of the study.
Each of these three measures of access can be used alone or in combination to
perform a spatial analysis. Each one sheds light on a different aspect of spatial access.
Measuring the distance to the closest park answers a different question than calculating
the amount of park area or creating park service areas. By performing a spatial analysis,
basic understanding of how accessible parks are to residents is gained. However, further
insight can be gained by pairing the spatial results with other data, so as to compare and
contrast how spatial access varies between groups. Spatial data can be combined with
many types of data to answer an endless variety of questions, for example some access
studies use spatial analyses to gain insight on how park access is related to physical
(Kaczynski et al., 2008; Koohsari et al., 2013) or mental health (Markevych et al., 2014;
Nutsford et al., 2013). This study, like many others, focuses on the differences in spatial
access between residents with different socioeconomic characteristics. The next section
discusses how the measures of access discussed above are joined with socioeconomic
data to perform equity analyses and the results found from previous studies.
Measuring Park Access: Protocols for Equity Analyses
When paired with socioeconomic data, the results from the access measures
discussed above can be statistically analyzed to help assess the equity of access to parks.
In this section I will discuss not only how previous studies have performed equity
analyses using spatial data and statistics, but also the results found using them. There will
be some emphasis on the measures and variables these studies used (both of which were
33
!
discussed in the previous two sections), so as to demonstrate how these selections may
affect the results, which is an overall aim of this thesis.
As mentioned previously, this study follows the example of previous studies by
defining equity as a traditionally disadvantaged socioeconomic group having better
spatial access than their privileged counterpart (Nicholls, 2001; Kabisch & Haase, 2014;
Talen & Anselin, 1998; Comber et al., 2008). A commonly used statistical test in park
access studies is the Mann-Whitney U test (Nicholls, 2001; Nicholls & Shafer, 2001;
Talen, 1997).
This is a nonparametric procedure for comparing two independent
samples. It is very similar to the commonly used parametric t-test for independent
samples. The Mann-Whitney U test uses ranked ordering to compare the distributions of
the two samples in order to determine if one sample is systematically higher or lower
than the other. In other words, the data from two samples is combined, placed in order,
and receives a rank. If one sample is systematically higher or lower ranked than the other,
to the extent to which it would have occurred by chance at most 5% of the time, it means
the samples are significantly different. This means that the null hypothesis (H0) of this
statistical test is that the distribution of both samples are not systematically lower or
higher than the other, with the alternative hypothesis (H1) being that the distribution of
one sample is lower or higher than the other (Corder & Foreman, 2014). Wading through
all the statistical jargon, the key point is that in an equity analysis, the way that equitable
access is determined using a Mann-Whitney U test is through the comparison of the
spatial access of disadvantaged and privileged groups. If the disadvantaged group is
found to have significantly better access, it is considered equitable. If it is found to have
significantly worse access is considered inequitable. Or if there is an insignificant
34
difference between the access of disadvantaged and privileged groups it is considered
equal access. In the case of the three measures of access described above, nearest park,
amount of park area, and park service areas, data can easily be split into two samples to
enable a Mann-Whitney U test. The following paragraphs will examine each of these
measures and discuss how Mann-Whitney U tests are used to determine socioeconomic
equity.
When measuring spatial access through the distance to the closest park, the data
can be split into quantiles, such as halves or quartiles, to compare the groups with the
shortest and longest distances to the closest park. The socioeconomic characteristics of
these two samples can then be statistically analyzed to determine if access is equitable for
disadvantaged groups. In a park access and method comparison analysis conducted by
Higgs et al. (2012), the authors used quintiles to perform equity analyses that compared
the distance to the closest park with the material deprivation (as measured using the
Townsend deprivation index based on unemployment, non-car ownership, non-home
ownership, and household overcrowding) of census tracts in Cardiff, Wales. This study
performed two equity analyses using Euclidean and network metrics so as to compare the
results. Interestingly, the two metrics produced different results when comparing the
distance to the closest park in the most and least deprived tracts. When the distance was
measured using the Euclidean metric, the distance was significantly shorter for the most
deprived tracts and longer for the least deprived tracts. When using network distance, the
most deprived tracts were still closer than the least deprived tracts to the nearest park, but
the difference was not significant. Using the previously stated definitions of equal and
equitable access, this study found the deprived tracts to have equitable access using
35
!
Euclidean distance, and equal access when using network distance. This study highlights
the importance of choosing appropriate variables and methods to measure access, and
even suggests the use of multiple methods to demonstrate stronger results and lack of
error (Higgs et al., 2012). Similar to Higgs et al. (2012) this Seattle case study uses
multiple methods in order to not only examine the equity of park access in Seattle, but
also the consistency of results between variables.
When measuring the amount of park area within a certain distance from a census
unit the results can be broken into the quantiles of those areas with the most and least
amount of park space. Talen (1997) created service areas around block centroids to
calculate the amount of park area within one and two miles in the cities of Pueblo,
Colorado, and Mason, Georgia. By splitting the results into quartiles, Talen was able to
compare the socioeconomic characteristics of the blocks with the best access and the
blocks with the worst access. Interestingly, with the use of a Mann-Whitney U statistic,
the author found that the two study areas produced very different results. In Pueblo,
Colorado, higher-income census blocks had significantly more access to park acreage,
whereas in Macon, Georgia, lower-income blocks had significantly more access to park
acreage. Additionally, census blocks with more nearby park area tended to be majority
white residents in Pueblo, but majority non-white residents in Macon (Talen, 1997).
Similar to Talen (1997), Boone et al. (2009) also found significant results by conducting
a park area analysis paired with race/ethnicity data in Baltimore, Maryland. However,
compared to Talen’s (1997) use of large network service areas, this study used much
smaller Euclidean service areas of only one-quarter mile to measure the amount of park
area within walking distance of block groups. Results showed that block groups in which
36
the majority of the residents were white tended to have more area of park within the
quarter mile threshold than block groups in which the majority of the residents were
African American (Boone et al., 2009). These studies show how inconsistent results of
equity analyses can be depending on the variables used and the case study location
characteristics. This is important to note, since there is no set standard variables for
measuring spatial access, it is somewhat difficult to compare results between case studies.
However, by implementing multiple methods in this Seattle case study, comparison is
more possible, but still somewhat difficult since each study area has different
socioeconomic demographics.
Lastly, the results from a spatial analysis looking at the service areas of parks can
be used in an equity analysis in a similar way. Instead of dividing the data using
quantiles, the socioeconomic data of the population inside the service areas is compared
with the population outside the service areas. For example, Nicholls (2001) used a halfmile threshold in her study using park service areas in Bryan, Texas. The author utilized
both Euclidian and network service areas as a means to compare the methods and
advocate for the use of network distance in spatial analyses. By implementing the MannWhitney U test, Nicholls performed an equity analysis by looking at age, race, mean
housing value, and renter-occupied housing (housing attributes are often used as a proxy
for income which is not available at the block group level). To the author’s surprise, the
results were very similar between the metrics. Regardless of the metric used, the
variables of high population density, low percentage of non-white residents, low housing
values, and high renter-occupied housing were all significantly associated with living
inside the half-mile service area. However age related variables produced mixed results.
37
!
The percentage of residents over the age of sixty-four was significantly higher inside the
service area when using Euclidean but not network service areas, whereas the opposite
was true for the percentage of residents under eighteen. Using the previously mentioned
definition of equitable and equal access, Nicholls found that across all variables there was
at least equal access to parks, and in many cases equitable access for the disadvantaged
groups. More specifically, this means that socioeconomic characteristics were distributed
at least equally between the area inside and outside of the half-mile threshold, and in
many cases a higher percentage of traditionally disadvantaged residents were found
inside the service areas than outside them (Nicholls, 2001). Similarly, Cutts et al. (2009)
performed a study in Phoenix, Arizona utilizing this access measure, but utilized a
quarter-mile Euclidean service area. By applying census data to the spatial results, the
authors found that the park service areas were positively associated with areas of high
Latino/a population as well as high African American population. However there was a
negative association between the service areas and the youth population. These results
indicate that although access was equitable for some traditionally disadvantaged groups
in this study, it was not true for all (Cutts et al., 2009). Using the same variables, Boone
et al. (2009) also used a quarter-mile Euclidean service area to compare the populations
inside and outside walking distance of parks. The authors used race and income data to
determine if there was equitable park access based on socioeconomic status. Results
showed that low-income and African American populations were significantly higher
inside the quarter-mile threshold than outside, indicating that these traditionally
disadvantaged groups did indeed have equitable access to parks (Boone et al., 2009).
38
Similar to the other access measures, using park service areas to measure access can be
affected by the variables chosen and the study area demographic.
Measuring access to parks is a growing and developing field of study. As this
section demonstrates, there are many ways of measuring the equity of spatial park access,
and the results are not uniform. However, all of the methods relate back to the spatial
relationship of people and parks. In a world that is growing more urban every year, the
importance of green spaces in an urban environment is more prominent than ever.
Studying this relationship helps gain further understanding of how cities can improve
spatial park access for their residents to improve quality of life. Through the use of GIS,
case studies can be performed as a way to not only further the understanding of patterns
of access and inequity, but also to test the accuracy and quality of the methods used.
STUDY AREA HISTORY AND DEMOGRAPHICS
This section is an introduction to the region, city, and park system of Seattle
Washington, so as to supply a larger framework and rationale for a case study. First the
history of the study area and the park system is discussed, as it is important to understand
the historical context that determined present park distribution and access. Secondly, an
in-depth look at present day Seattle and the park system is presented, so as to demonstrate
how the demographics are changing and how that could affect park use and access.
History of Seattle and its Parks
Seattle is the largest city in the northwestern United States. It is located in western
Washington, with Puget Sound and Olympic Mountains to the west, and Lake
39
!
Washington and the Cascade Range to the east (See Figure 1). The region was shaped
during the Pleistocene, when a series of continental glaciers advanced and retreated over
the area, carving it into the hilly terrain present today. The climate is generally mild and
wet with a dry period in the summer. This variation is due to the seasonal changes in the
high and low-pressure areas in the North Pacific Ocean. During most of the year moisture
enters the Puget Sound basin from the ocean, and results in precipitation when it meets
the mountain ranges. In summer the pressure shifts and comparatively dry and mild air
enters the basin, warming as it moves inland to the Cascades (Kruckeberg, 1998).!
Figure 1: Regional map of Seattle area
40
Although details are not precisely known, it appears that humans arrived to the
area around 8,000 years ago, as the last of the glaciers receded (Kruckeberg, 1998). With
an abundance of natural resources, the Puget Sound basin has been home to indigenous
peoples for thousands of years. In the strip of land that was to become Seattle, lived the
Duwamish people (Thrush, 2007). Europeans started exploring the area in the late 18th
century, most notably George Vancouver in the 1790s who gave the inlet the name Puget
Sound (Klingle, 2007). There are mixed accounts of how Seattle came to be, but the
textbook version is that in November 1851, Arthur Denny and a party of twenty-four
people landed in what is now West Seattle where they were greeted by Chief Seeathl
(also known as Sealth) of the Duwamish people. Within a year the Denny Party crossed
over to the shores of Elliot Bay were they built a settlement and named it after the chief
who had greeted them, Seattle. It was not much longer until the first sawmill was built in
the bay by Henry Yesler, bringing what was to become the dominant industry into the
area (Klingle, 2007; Thrush, 2007). This account glosses over the horrible decimation of
the indigenous people of the area. Although perhaps a less violent transition of power
than other areas of the United States, it is part of Seattle’s history that is often brushed
over yet somehow appropriated and romanticized by the city today. Even from the
earliest days of the settlement, place names and monuments were named after a mix of
both members of the founding Denny Party and the contemporary indigenous people (for
example Boren Avenue, and the Leschi neighborhood) (Klingle, 2007; Thrush, 2007).
Over the next few decades the settlement developed into a city. By 1880 the
population had rapidly grown to 80,000, despite a large fire in 1889 that burnt down all of
downtown. Over the next ten years the population grew to 240,000 residents. Lumber and
41
!
coal were dominant industries, which along with city expansion soon led to further
changes in the landscape: tidal flats of Elliot Bay were filled in, the delta of the
Duwamish River was dredged, and a canal connecting Lake Washington to Puget Sound
was dug, permanently cutting the city in half with a waterway (Klingle, 2007).
It was during this same time that Seattle’s first public park was created and the
municipal park system began to take shape. In 1884 Denny Park was created, becoming
Seattle’s first publicly owned park. Over the next several years a handful of parks were
created. Soon the city saw a need for a comprehensive park plan in order to create and
integrate a park system into the city before economic competition, population increase,
and further city expansion set limitations. In 1903 the city commissioned the Olmsted
Brothers landscape architecture firm of Brookline, Massachusetts to design a system of
parks throughout the city of Seattle. The Olmstead Brothers were the sons of the famed
Fredrick Law Olmstead, whose landscape architecture firm designed Central Park of New
York City in 1859. His son, Fredrick Law Olmstead Jr., and his orphaned nephew whom
he adopted, John Charles Olmstead, continued his legacy as landscape architects. John
Olmstead took the lead on the Seattle project. Arriving in the spring of 1903, He was
unimpressed by the city’s already established parks’ ridged urban feel. Olmstead saw a
vast untapped potential in the city’s natural landscape and used it to design a park system
that took into account the natural topography of the land. The proposed system was a
series of large parks joined together by parkways and boulevards along the shorelines,
waterways, and ridgelines (See Figure 2) (Dooling, Simon, & Yocom, 2006; Klingle,
2007).
42
Figure 2: Historical map of Seattle parks
(Seattle Board of Park Commissioners, 1909)!
43
!
There were many controversies and compromises in the creation of the Olmsted
designed park system. Most of the issues were a product of the conflict between
Olmstead’s vision and the functionality of a growing city. Many conflicts arose between
the city engineer and Olmstead, fighting over where parks, sewer lines, water mains, and
roads should be located. Olmstead was an advocate for preserving what was left of the
old-growth forests, and was able to convince the city to detour planned arterials away
from some of them. Additionally, from very early on Olmstead expressed concern over
the inequity of park distribution, believing that the low-income and non-white
communities needed improved park access. For example, Olmstead noted a distinct lack
of park space in the southern part of the city, which led to the purchase of a large tract of
land on the southwest shore of Lake Washington. There were many compromises made
over the park plan, but during the more than thirty years that John Olmstead worked on
the Seattle park system, the key elements of large parks and parkways did come to
fruition. Some of the earliest parks include Green Lake Park, Ravenna Park, Washington
Park Arboretum, and Seward Park, with sweeping boulevards that follow the terrain like
footpaths built adjacent to them, perhaps most famously Lake Washington Boulevard
(Dooling et al., 2006; Klingle, 2007).
Since the days of Olmstead, the Seattle park system has continued to grow and
develop. Like many cities across the United States, the Great Depression, and the World
Wars affected Seattle’s economy, including park acquisition and upkeep. As the city
continued to grow, major thoroughfares were built, cutting through the city and creating a
disconnect between parks. In general, Seattle voters have been supportive of the city park
system, however, in the 1950s there were a handful of propositions to fund the continued
44
development and maintenance of the park system that were not supported. This lack of
funding may be partially to blame for an increased unease between the citizens and the
city government in the 1960s due to inequality in the upkeep and distribution of the city’s
infrastructure and parks. By the 1970s, new propositions giving large amounts of funding
to parks had passed, as well as a federal program granting surplus federal land to the local
government. Over the next decade more than a thousand new park acres were acquired to
create sixty-four new parks, including the massive Discovery and Magnuson Parks, both
ex-military bases. As large tracts of vacant land became more obsolete and expensive, the
parks department became creative with park development by converting already
developed land into parks. A famous example is Gas Works Park, a former gas
manufacturing plant that incorporated the industrial features into a unique park landscape
that became one of Seattle’s landmark parks in 1975. Additionally, an abandoned railroad
line along the shore of Lake Washington was converted into a thirty-five mile long
bicycle and pedestrian route, furthering John Olmstead’s objective of a connected park
system (Dooling et al., 2006).
In most recent history, the voters of Seattle passed two large levies. In 2000
Seattle resident’s approved the Pro-Parks Levy that granted the parks and recreation
department nearly two-hundred million dollars for new parks, upgrades, and maintenance
(Dooling et al., 2006). In 2008, a similar levy (of $145 million) passed, demonstrating the
continued support of the park system (SPR, 2014).
45
!
The Present and Future of Seattle and its Parks
Since the days of its founding, Seattle has continued to grow dramatically.
Residing in King County, Seattle has an area of around eighty-four square miles. In 2010
Seattle’s population was over 608,000, with an estimate of over 686,000 in 2016 (See
Table 1 for 2010 Census Summary). In terms of population density, the city’s population
grew from around 7,200 to over 8,100 people per square mile in 2010 and 2016
respectively. Between 2000 and 2010 Seattle experienced an 8% population increase,
Table 1: Demographics of Seattle Washington from the 2010 Decennial Census!
46
however from 2010 to 2016 Seattle’s population grew by grew by over 12% (USCB,
2010b; City of Seattle Department of Planning and Development (DPD), 2011; Office of
Financial Management Forecasting & Research Division (OFM), 2016). The majority of
the population is white (66%), with Asian (14%), and African American (8%)
populations being the largest minorities. Although Seattle is still a predominantly white
population, the population growth rate between 2000 and 2010 was nearly double for
non-white race/ethnicities versus that of whites (13% and 7% respectively). Adults (ages
18 to 64) make up the bulk of the population (74%), while children (under 18) outnumber
older adults (65 and older) (15% and 11% respectively), although the percentage of older
adults in expected to grow in the coming decades as baby boomers age (DPD, 2011; SPR,
2014). The household median income has grown rapidly in the last decade, growing by
$30,000 in ten years from 2005 to 2015, with a nearly $10,000 increase between 2014
and 2015 (USCB, 2005; 2014; 2015).
As of 2016, the Seattle park system consisted of over 485 parks (including
designed and developed parks as well as natural green spaces), with over twenty-five
miles of boulevards and 120 miles of trails, totaling more than 6,400 acres of parkland or
about 12% of the total city area (SPR, 2016). The Trust for Public Land (TPL), a
nationwide non-profit organization dedicated to the creation and protection of public
land, found that in 2016 Seattle had 9.9 acres of parkland per 1,000 residents, far less
than the median for the one hundred largest cities in the United States (13.1 acres).
However Seattle’s acreage is slightly over the national average of 9.6 acres per 1,000
residents found by the National Recreation and Park Association (NRPA), a non-profit
organization dedicated to the advancement of public parks that compiled the data from
47
!
nearly one thousand park systems across the country to create an extensive and publicly
available dataset. Using NPRA’s interactive dataset to compare medians, Seattle is below
average for park systems in the Pacific Northwest (which had a median of 12.9 acres per
1,000 residents), cities larger then 250,000 (which had a median of 12.2 acres per 1,000
residents), and park systems with over 3,500 acres of park area (which had a median of
17.2 acres per 1,000 residents). However Seattle has more park acreage than most
jurisdictions with a similar population density of more than 2,500 people per square mile
(which had a median of 7.7 acres per 1,000 residents). Additionally, TPL found that more
money was spent on parks per resident in Seattle than any of the other cities in their one
hundred city study (Trust for Public Land (TPL), 2016; National Recreation and Park
Association (NRPA), 2017). I mention these statistics to demonstrate how Seattle
compares with national and regional averages, and to provide some insight on the current
status of the park system. Seattle is nicknamed ‘the emerald city’ but it does not actually
have more park acreage than its comparable systems. However, the fact that the SPR
spends a large amount of money per resident demonstrates that parks are a priority for the
city, and indicates that money might not be a limiting factor in the system’s ability to
adapt for population increase and densification.
As part of Seattle’s adaptation to its continued population growth and
densification, Seattle Parks and Recreation (SPR) created the Parks Legacy plan, named
to honor the heritage of the park system created by Olmsted. This document lays out not
only the current status of the park system, but also its future goals and strategies.
Including overviews of how Seattle parks affect residents as well as the environment, this
report provides a closer look at how the benefits discussed early in this manuscript are
48
present in Seattle. Additionally, the plan contains information on the socioeconomic
characteristics of the city and how to better serve the changing population (SPR, 2014).
In the Park Legacy Plan, SPR has the core values of access, opportunity, and
sustainability. The plan highlights the importance of having a park system that responds
to emergent needs and provides access and opportunities for park use. Additionally, the
plan emphasizes achieving the outcome goals of “healthy people, a healthy environment,
financial sustainability, and strong communities” (SPR, 2014, p.1). SPR strives to
promote physical and mental health in its jurisdiction. At the time the plan was written,
over half of King County’s adult population was overweight or obese, highlighting the
importance of improving public health as a central goal. Additionally, Seattle parks
benefit residents and the environment by containing over 600,000 trees. This contributes
greatly to a vast array of benefits previously discussed including improved air quality,
carbon sequestration, rainwater retention, climate regulation and wildlife habitat. Seattle
parks strengthen communities and contribute to the social well being of the residents. By
supplying places to gather and build relationships, as well as through athletic facilities,
youth employment, and volunteer programs, SPR promotes stewardship and a sense of
belonging in neighborhoods across the city. Furthermore, as a byproduct of achieving
their goals, the Seattle park system benefits the residents and the city by saving millions
of dollars each year. In a study conducted by TPL, it was found that Seattle residents save
over sixty-four million dollars annually in saved medical expenses through increased
physical activity in parks. TPL estimated that the city government saved over twelve
million dollars through the stormwater management, pollution mitigation, and
community cohesion. In addition to money saved, money is earned for residents and the
49
!
city through increased property value and related taxes, as well as the tourism industry
(TPL, 2011). These examples demonstrate how SPR is working toward and meeting their
goals, as well as highlighting the benefits of parks in Seattle.
As part of the Park Legacy Plan, SPR conducted a phone survey in 2012 to gain
insight on how to better serve the community. Conducted by phone, the Park Legacy
Survey had four hundred adult respondents that reflected the overall demographics of
Seattle. The results help to shed light on how residents use the parks, and how the park
system can improve.
When pairing the results with race and ethnicity, it was found that Hispanic and
respondents who self-reported their race as ‘other’ were the race/ethnicity categories with
the highest percentage that visited a park at least once a week (both 70%), followed by
African American respondents (57%), and whites (52%), while Asian respondents
reported the smallest percentage (28%). When asked about visitation and use of
playgrounds, athletic fields, community centers, recreation programs, and picnic shelters,
non-white residents reported a higher rate of weekly use than whites in all categories. The
majority of white respondents chose “exercise and fitness” as a top reason to visit parks,
whereas a majority of non-white respondents chose “socializing with family and
neighbors” (SPR, 2014, p.12-13). This demonstrates that although the majority of Seattle
is white, the park system is used at a higher rate by minority races and ethnicities,
demonstrating a need to assess park access for those communities to ensure that access is
equal or equitable.
Additionally, income was determined by the survey to be a strong predictor of
park visitation. Results showed that residents making less than $50,000 a year are much
50
less likely to visit parks on a weekly basis than those making over $100,000 (39% and
68% respectively) (SPR, 2014). These results are somewhat unexpected given the
relationship of poverty and race in the United States. Nationally, census data shows that
poverty rates are lowest for white people and people of Asian decent, with all other races
experiencing a higher rate of poverty. In Seattle, a 2015 estimate shows that fewer than
10% of white people are below poverty level, while all other race categories ranged from
11.7% (Asian) to around 25% of African Americans, Hispanic/Latino, and residents
marking ‘other’, with an average of 16% for all non-white residents (USCB, 2015).
Given the SPR survey results finding that non-white residents tend to use parks at a
higher rate, it was surprising to find that income did not follow that same trend. This
discrepancy brings to light the issue of access as an interdisciplinary issue, and whether
spatial access plays a role or if other factors are limiting use.
Survey results showed that age also played a role in park use. Respondents aged
35 to 54 were the most frequent park users, with two thirds (66%) visiting a park weekly,
while the same is true for less than half (42% and 41% respectively) of younger adults
(aged 18 to 24) and older adults (aged 55 and older). Because the senior population is
projected to grow, it is important for their interests and needs to be taken into account.
Likewise, it is important to take the needs of children and families into account since the
survey indicated that households with children are more likely to visit parks on a weekly
basis than households without children (73% and 41% respectively)(SPR, 2014).
Although access to parks is important for all age groups, it is especially important for
children and older adults, who benefit greatly from physical activity and social
interaction (Bedimo-Rung et al. 2005; Sherer, 2003).
51
!
This section has shown that Seattle is a strong candidate for an in-depth
accessibility study. It is an extensive system, with a history of extensive park planning as
well as wide public support. The parks and recreation department has a main goal of
promoting access to its parks. This case study creates a benchmark for Seattle park
accessibility and illuminates the areas in need of improvement. The changing
demographics of Seattle make it an interesting study area. Results from the Park Legacy
Survey helped to indicate socioeconomic characteristics such as race, age, and income
variables that were examined in the equity analysis of this Seattle case study, enabling the
comparison between the two sets of results. All of these factors provide a foundation for
an accessibility study and a subsequent evaluation of GIS methods, yet none has been
completed up until now. Additionally, this case study contributes to the body of park
access literature. Case studies are an important part of gaining knowledge about park
access and how to measure it. This case study provides insight on the subject in an area of
the United States that had not yet been examined. Using the results from previous
research to determine the best ways to test methods in a new study area helps to
strengthen the methods and to set a standard for measuring access. A study in Seattle
helps gain insight on the city’s park system and how it could improve, as well as for the
research community in order to further knowledge of park access methodology.
52
CHAPTER 3: METHODS
The aim of this study was to determine if residents of Seattle, Washington,
regardless of location or socioeconomic characteristics, live within an equitable distance
to similar amounts of public park acreage and whether the methods used affect the
outcome of the results. Building upon the methods and studies discussed above, I
completed this spatial accessibility and socioeconomic equity analysis case study in
Seattle, Washington by combining several socioeconomic variables and spatial access
measures in order to not only contribute new empirical data to the growing body of
research, but also provide new insight on which methods are most successful and
appropriate for measuring access. Specifically, I completed this spatial analysis with
geographic information systems (GIS) software using both Euclidean and network
distance metrics, as well as census data relating to race/ethnicity, housing occupation
status, age, and population density at both block and block group census unit aggregate
levels. In this section the specific details of the methods used in this study will be given.
First I will explain were I gathered data and how I prepared it for analysis. Then I will
give details of the spatial analysis I completed, with details of the variables I chose, and
how I measured spatial access. The last section is focused on the statistical analysis, and
how it was used to complete an equity analysis and method comparison analysis.
DATA ACQUISITION AND PREPARATION
I used a variety of data sources in this case study. I obtained data regarding people
from the United States 2010 Census (USCB, 2010a). All data was obtained at both the
53
!
block and block group aggregate level. Only data available at both levels were used in
order to provide direct comparison. The attribute data used was number of housing units,
occupied housing units, vacant housing units, owner occupied housing units, renter
occupied housing units, households with residents under 18, households without residents
under 18, households with residents over 64, households without residents over 64, total
population, white population, and all other race categories were combined to create a
non-white population category. I used the categories regarding housing occupation as a
proxy for income, since income information is not available at block and block group
level (following the example of Nicholls, 2001).
I obtained spatial data in the form of shapefiles, a common type of spatially
referenced data file, from two sources that are both available for use to the public. I
gathered data regarding the land and waterways of the area, as well as the block and
block groups from the King County GIS data portal (King County, 2017). The City of
Seattle open data portal provided the data on the Seattle municipal park system as well as
the road and trail networks (City of Seattle, 2017).
I carried out the spatial analysis using ArcGIS Desktop (Version 10.5) with the
Network Analyst extension, an application package produced by Esri, a mapping
software company. The spatial data was loaded into the program and projected using a
consistent geographic projection system and coordinate system. I joined census data to
the spatial data with the result that each individual spatial block and block group was
associated with the specific socioeconomic data for that location.
54
SPATIAL ANALYSIS
I used three methods to measure spatial access in this analysis. These methods,
which I discussed extensively in the literature review chapter, measure the distance to the
nearest park, the park area within a certain distance, and park service areas. I completed
each of these measures using Euclidean and network distance metrics as well as using
both block and block group census aggregation units, resulting in a total of twelve GIS
based spatial analyses (see Table 2 for full list). The metric used, Euclidean or network,
affected what specific GIS tools I used to perform the measurements for the spatial
analysis, but the census unit did not affect the protocol, only the amount of data being
processed. For all spatial analyses, the block groups were represented by population
weighted centroids (a publicly available dataset created by the U.S. Census Bureau), and
the blocks by geographic centroid (which were created using the ‘Mean Center’ tool). As
Table 2: List of spatial analyses performed using combinations of
census units, metrics, and access measures
55
!
I discussed previously in the literature review chapter, geometric centroids use the
coordinates of the polygon to find the average center point of the shape, and population
weighted centroids are the average location of the population within an area (Bolstad,
2012). Due to blocks’ relatively small size, population weighted centroids are not
available and the difference between geographic and population weighted centroids
would be minimal (See Figures 3-5 for examples of census units, all of which show the
same extent of a neighborhood in Seattle that selected only as an example to demonstrate
the differences in census unit). Additionally, due to the high number of unpopulated
blocks (mostly in highly industrial and business areas), I excluded these from the analysis
in order to better represent resident access; all block groups were populated and included
in the analysis. Parks were represented by their boundaries. In ArcGIS the network metric
can only measure between points, as opposed to the Euclidean metric, which can measure
between any combination of points, lines, and polygons. Due to this restraint, Euclidean
measurements used park boundary lines, while network measurements used points that I
created at 50-foot intervals along the boundary lines of the parks (see Figure 6 for
example of boundary points). The network used for measuring network distance included
the streets and paved trail networks, but did not include freeways or railways. These
inclusions and exclusions were used so as to best represent residents travelling via
walking, or even bicycle, but not motor vehicle. As seen below in the specific protocols
for each measure, I chose short threshold access distances to also represent non motor
vehicle transportation.
56
Figure 3: Close-up of Seattle neighborhood showing size difference of blocks
and block groups
!
57
!
58
Figure 5: Close-up of Seattle neighborhood showing
blocks with geographic centroids!
!
Figure 4: Close-up of Seattle neighborhood showing
block groups with population weighted centroids
Figure 6: Close-up of Seattle neighborhood showing example measurements from
block group centroids to the nearest park boundary or boundary point using network
and Euclidean metrics.
!
!
Park Proximity
Measuring to the nearest park was relatively simple (see Figure 6 for example).
When using the Euclidean metric, the ‘Spatial Join’ tool with the ‘nearest’ match criteria
was used. This tool used the specified starting census unit centroid and found the closest
park polygon boundary, calculated the Euclidean distance between the points, and
repeated this for every census unit centroid. When using network distance, the Network
Analyst extension toolset was used to perform a ‘Closest Facility’ analysis. This analysis,
similar in theory to the Euclidean method, took every census unit centroid and measured
to the nearest park boundary point. The main difference is that instead of measuring a
straight line, it measured along the road and trail network. The results of these analyses
were displayed in the form of tables with an entry for each census unit with all of the
associated census data and a new distance measurement.
59
!
!
Figure 7: Close-up of Seattle neighborhood showing an example of the area within a
quarter mile of a block group centroid measured using a Euclidean and network service
areas. Demonstrates the size difference, and the effect on included park area.
Amount of Park Area
The second measure of accessibility I used was the amount of park area within a
certain distance from census units (see Figure 7 for example). For the analysis, I used a
threshold distance of a quarter mile to represent a short walking distance. When using the
Euclidean metric, I used the ‘Buffer’ tool to create a service area polygon around each
census unit centroid using the quarter mile distance. Since the input was a point, the
service area was essentially a circle with a quarter mile radius and the census unit
centroid at the center. Using these service area polygons as the input, I used the ‘Spatial
60
Join’ tool with the ‘intersect’ match criteria. This tool summed the areas of the parks
whose boundaries fell within the Euclidean service areas for each census unit centroid.
Once again, using the network metric was not very different in theory. I used the
Network Analyst extension toolset to perform a ‘Service Area’ analysis. Essentially what
this did was similar to the creation of the Euclidean service areas. Starting at each census
unit centroid, a quarter mile was measured out along all accessible streets and trails and
the endpoints were connected to create an irregular polygon. I then used ‘Spatial Join’ in
the same way as the Euclidean protocol to return a summation of park areas for each
census unit centroid.
Park Service Area
The final measure of accessibility I utilized looked at the residents inside and
outside the service areas around parks (see Figure 8 for example). For the analysis, I used
the same threshold distance of a quarter mile used in the previous measure. When using
the Euclidean metric, I created service area polygons around each park by using buffers
to measure a quarter of a mile out from around the park boundaries. Using the ‘Dissolve’
tool, I combined the buffers into a single service area polygon for all parks. Using the
‘Select by Location’ tool with the ‘within’ criteria, I selected census unit centroids that
were within the quarter mile service area polygon and, using the ‘Field Calculator’ the
units were marked as ‘inside’. I used the ‘switch selection’ tool to then select all of the
centroids that fell outside of the service area polygon, which were marked as ‘outside’.
Essentially this split the census data into two groups: those inside and outside a quarter
mile of parks. This is unlike the previous two measures, were a unique distance and area
61
!
Figure 8: Close-up of Seattle neighborhood showing park service areas using Euclidean
and network service areas. Demonstrates the size difference and effect on total units
included inside area.!
value were calculated fro each unit. This measure is binary with only ‘inside’ or ‘outside’
value options. For network distance, I again used the Network Analyst extension toolset
to perform a ‘Service Area’ analysis. Due to the tool inability to use polygons as an input,
the 50-foot interval boundary points each acted as an origin point. This created thousands
of service area polygons, which I then dissolved into a single park service area polygon.
Identical to the Euclidean methods, the census unit centroids were divided into those
inside and outside of the service area.
62
I used each of these three access measures in a total of twelve spatial analyses,
using every combination of block and block group census units as well as Euclidean and
network metrics (See Table 2 for full list). The tables associated with these spatial results
were then exported to Microsoft Excel for reformatting.
STATISTICAL ANALYSIS
The statistical analysis consisted of two parts. In the first part I analyzed the
socioeconomic data to perform an equity analysis in which I compared how the spatial
results differed between groups. In the second part I analyzed the use of Euclidean and
network metrics as well as the use of block and block group census units to determine if
the spatial results differed between variables. The Mann-Whitney U test, which I
described in the literature review chapter, was used for all statistical analyses. IBM SPSS
Statistics was the software used to perform the statistical analysis. The results of the
Mann-Whitney U test, which uses ranked ordering to compare the distributions of two
independent samples, are found to be significant only when the ranking of the two data
samples are distributed differently enough that they would have occurred by chance at
most 5% of the time (having a p-value of less than 0.05). To determine which of the two
samples had a higher or lower distribution, the mean of the ranks of the two samples were
used. I used test to complete two series of analyses. First an equity analysis was
completed, in which the socioeconomic characteristics of the census units with the best
and worst spatial access were compared. Secondly performed a method comparison
analysis to compare the use of census unit and metric variables to determine if they
affected the spatial results.
63
!
Socioeconomic Equity Analysis
First I completed the socioeconomic equity analysis. In Excel, the raw
socioeconomic data was converted into percentages, following the example of Talen
(1997), Nicholls (2001), and Nicholls and Shafer (2001). This is to normalize the data so
as to better represent each socioeconomic group. The socioeconomic variables I used for
the statistical analysis were (i) percent vacant housing units, (ii) percent renter occupied
housing units, (iii) percent of households with residents under 18, (iv) percent of
households with residents over 64, (v) percent non-white population, and (vi) population
density. Following the example of Nichols (2001) I used housing variables as a proxy for
income due to census data on income not being available at the block or block group
aggregate level. Due to Seattle having a large white majority, I chose to combine the data
for non-white residents in order to create a larger representation. This was following the
example of Nicholls (2001) and Talen (1997). For the nearest park, I split the data into
quartiles by distance to park. The top quartile, the census units with the closest parks, was
compared to the bottom quartile, the census units with the farthest parks. Similarly, for
park area the data was split into quartiles by amount of area. The upper quartile, the
census units with the largest amount of park area, was compared to the bottom quartile,
the census units with the smallest amount of park area. For the residents around parks, the
data was divided into census units inside and outside the quarter mile threshold. Each
combination of spatial measurement, census unit, and metric was used in order to
facilitate a comparison. This amounted to a total of twelve rounds of Mann-Whitney U
tests each consisting of six individual socioeconomic variable comparisons (See Table 3
for list of all study variables). In each test I determined for each of the six socioeconomic
64
variables whether the two samples (near or far, small or large, inside or outside) were
distributed significantly different, and which sample was distributed higher or lower.
Equitable access was determined to have been achieved only when the disadvantaged
groups, young residents, older residents, non-white residents, home renters, areas with
vacant housing, and high population density were distributed significantly higher in the
samples representing closer parks, more park area, and inside park service areas. When
results were not significant it indicated the two samples had similar park access. Only
when results showed a traditionally privileged group having significantly better access
was inequity of park access claimed.
Table 3: List of all variables used in study
Method Variable Analysis
After completing the equity analysis, I performed several follow-up MannWhitney U tests to further evaluate the differences between census unit and metric
variables. A series of eight analyses compared Euclidean with network results and block
with block group results to see if the metric or census unit used had a significant impact
on the spatial results. When comparing metrics, the data resulting from the spatial
65
!
analyses using both block and block groups were combined by the metric used. The
inverse was true when comparing census units, meaning Euclidean and network data was
pooled by census unit used. For each of the three measures of access, I ran a separate pair
of tests, one for metric, one for census unit. When looking at park proximity, I compared
the distance measurements found using each variable. Similarly, for park area the amount
of area was used in the analyses. However, because the spatial analysis of the service
areas around parks simply split the census units into groups, it did not produce a unique
value for comparison. To enable a Mann-Whitney U test, I analyzed the data in two
separate tests using data from inside and outside the quarter mile threshold. I used
population density as the variable to determine if there was a difference between the data.
Population density was used due to it being normalized to combine block and block
group data.
The statistical analysis consisted of a total of 80 Mann-Whitney U tests, the
results of which will be discussed in the next chapter.
66
CHAPTER 4: RESULTS
I will present the results in two parts. In the first I will report the findings of the
socioeconomic equity analysis using census blocks and network metrics. I selected these
variables for this more in-depth equity analysis due to the small size of blocks best
representing the spatial distribution and socioeconomic diversity of households, and the
network metric more accurately representing actual travel distances. There will be no
comparisons of methods, just the results of whether access to parks was found to be
equitable for disadvantaged groups using these variables. The second part will be a
comprehensive account of how census unit and metric variable selection affected the
results of the spatial and equity analyses. This section will give detailed results of how
each method variable affected the significance of the spatial and statistical results. I chose
to present the results in this order so as to first describe in-depth the equity results using
what variables I think are the best option based on previous studies, and then show how
results using other variables compare in terms of methodological implications. In this
chapter I focused on presenting the most important results found in this case study, it will
be followed by a discussion of what the implications of the results are in the subsequent
chapter. For reference, maps displaying the spatial distributions of socioeconomic
variables are available in the appendix but will not be discussed in the text.
PARK ACCESS AND SOCIOECONOMIC EQUITY RESULTS
Results of the equity analysis using blocks and the network metric show that
traditionally disadvantaged groups have equitable access to parks in many cases. This
67
!
analysis specifically looked at the variables for vacant housing units, renter occupied
housing units, non-white residents, households with residents under 18, households with
residents over 64, and population density as disadvantaged groups. Results will be
reported first by looking at the overall patterns of the analyses followed by more detailed
results within each spatial analysis used, starting with closest parks, followed by access to
park area, and lastly park service areas. The results of the Mann-Whitney U tests have
been compiled in Tables 4 and 5.
Overall Patterns
Overall, most of the traditionally disadvantaged groups examined in this Seattle
case study live within an equitable distance to similar amounts of public park acreage
(see tables 4 and 5 for details). The percent of vacant housing was the only
socioeconomic variable found to be significantly equitable by all three measurements.
Percent renter occupied housing and population density were found to have equitable
access as measured by park proximity and park service areas but only equal access to
park area. Percent non-white residents was equitable in the amount of park area as well as
park service areas, but only equal in the distance to the closest park. The variables
regarding age had the most mixed results. Percent of households with residents over 64
was found to be equitable in park area, equal in distance to the closest park, and
inequitable in park service areas. Percent of households with residents under 18 was
equal in in park area, but inequitable in distance to the closest park and being within park
service areas. These results indicate that Seattle has equal and equitable access to parks
for most disadvantaged residents, but that the youth and older residents do not have as
68
good of spatial access to parks compared to the general population. In the following
sections the detailed results of each access measure will be presented in order to gain a
deeper understanding of the equity of park access for residents in Seattle.
Park Proximity
When measuring access through park proximity, the median distance from block
centroid to park boundary was 875 feet, less than a fifth of a mile (N=8524; SD=762.7
feet). The data ranged from zero (meaning adjacent to the park boundary), to almost 7900
feet, just under one and a half miles. The blocks in the quartile with the closest parks
were a median of 221 feet away from parks (N=2131; SD=157.4 feet); whereas the
blocks in the quartile with the farthest parks were a median of over 1800 feet away from
parks (N=2131; SD=624.1 feet). See Figure 9 for a map displaying results of the spatial
analysis.
The results of the Mann-Whitney U-tests showed the variables percent vacant
housing, percent renter occupied housing, and population density had equitable access,
meaning they had significantly higher ranks in blocks in the upper quartile than the lower
quartile of access (p<0.00). Race/ethnicity demonstrated equal access; meaning blocks in
the upper and lower quartiles of park proximity had insignificant rank differences in
percentages of white and non-white residents (p=0.17). Older residents were also found
to have equal park access, with blocks having an insignificant rank difference in
percentages of households with residents over the age of 64 in the upper and lower
quartiles (p=0.69). The percentage of households with residents under the age of 18 was
the only category to result in significantly inequitable access (p<0.00). In other words,
69
!
blocks with a higher percentage of households with residents under 18 were significantly
more likely to be in the lower quartile than the upper quartile of access.
Amount of Park Area
Measuring access by calculating the amount of park space within a quarter mile of
block centroids resulted in a median value of 0.81 acres (although the mean was much
higher at 15.3 acres; N=8524; SD= 42.8). The data ranged from zero acres to 517.7 acres
of park area within a quarter mile of blocks. The blocks in the quartile with the most
amount of park area within a quarter mile had a median of 32.3 acres (N=2131;
SD=70.5). However the blocks in the quartile with the least amount of park area had a
median of zero, due to all blocks having no park area within a quarter mile (N=2131;
SD=0). See Figure 10 for a map displaying results of the spatial analysis.
Percent vacant housing, percent non-white residents, and percent households with
residents over 64 were shown to have equitable access in the results of the MannWhitney U-tests. The ranks of all three variables were significantly higher in blocks in
the upper quartile than the lower quartile of park area (p<0.00). The ranks of renter
occupied housing, household with residents under 18, and population density were found
to be insignificantly different between the quartiles with the most and least amount of
park area, demonstrating equal access (p=0.67; p=0.57; p=0.57). None of the variables
were found to have significantly lower ranks in the analysis, meaning no inequitable
access was found.
70
Park Service Area
When looking at the residents inside and outside of the quarter mile park service
areas, about 70% of blocks were considered inside. It was found that a median of 1245
residents lived within the service area around each park (N=432; SD=1643.9). The
populations ranged from zero residents to over 16,000 residents, and the population
densities ranged from zero to over eight people per square foot of park space. The median
population density for blocks inside the service area was 14.3 residents per acre (N=5948;
SD=24), while the blocks outside had a median population density of 12.7 (N=2576;
SD=13.5). See Figure 11 for a map displaying results of the spatial analysis.
The results of the Mann-Whitney U test indicated a significant difference between
the inside and outside blocks in all six of the variables tested. Equitable access was found
for percent vacant housing units, percent renter occupied housing units, percent nonwhite residents, and population density (p<0.00). Households with residents under 18 and
households with residents over 64 were both found to have significantly inequitable
access (p<0.00).!
71
!
72
!
Table 4: Results of equity analysis using network distance and block units. Green indicates equity with a significant p-value, red
indicates inequity with a significant p-value, yellow indicates equality with an insignificant p-value
73
!
Table 5: Results of equity analysis using network distance and block units. Values significant at p < 0.05, mean
rank and median used to determine which sample is distributed higher. ‘Best Access Samples’ refer to quartiles
with the closest parks, largest amount of park area, and inside the park service areas. ‘Worst Access Samples’ refer
to the quartiles with the farthest parks, smallest amount of park area, and outside of the park service areas.
Figure 9: Seattle with the results from park proximity spatial analysis using
blocks and network distance. Results are split into quartiles,
with dark color indicating closer access.!
74
Figure 10: Seattle with the results from the amount of park area spatial analysis using
blocks and network distance. Results are split into quartiles, with dark color indicating
more park acreage within a quarter mile.!
75
!
Figure 11: Seattle with the results from park service area spatial analysis using
blocks and network distance. Results are split into blocks inside and outside
the quarter mile threshold.!
76
METHOD VARIABLE ANALYSIS RESULTS
Results of the variable analysis demonstrate that in many cases both the census
unit and metric have an effect on the significance of results. In this analysis I specifically
compared the use of blocks with block groups and the use of Euclidian with network
metrics through comparing spatial results in comparative Mann-Whitney U tests, as well
as by completing an equity analysis for each combination. Results will be reported on
first by looking at some overall trends within each socioeconomic variable, and then in
more detail by the spatial analysis used. The results of the spatial analyses are
summarized in Table 8, while the Mann-Whitney U tests are shown in Tables 6 and 9.
Overall Patterns
Trends for each socioeconomic variable emerge when looking at the results of all
twelve equity analyses (See table 6 for equity results). Vacant housing units had the most
consistently equitable results, with eleven tests demonstrating significant equitable access
for census units with higher vacancy rates. This indicates that in this case study, variable
selection did not play a large role in results for this socioeconomic category, or that the
relationship was so strong that the significance was nearly consistent despite the
differences in variables.
Two variables, renter occupied housing and households with residents under 18,
remained consistent within the three access measures with park proximity and park
service area being significant, and park area being insignificant. These results indicate
that for these two categories metric and census unit variable selection did not have a huge
influence on the results, but that how access is measured can produce different results.
77
!
Meaning that just because there is equitable access by one measure does not mean there
will be equitable access measured in all.
The other three socioeconomic categories yielded more mixed results. Non-white
residents and households with residents over 64 were consistent in their results for
distance to the closest park, but produced mixed results for the amount of park acreage
and park service areas. Lastly, population density yielded mixed results in all three access
measures.
However it is important to note that when the significance of the results in a
single access measure were not consistent within a socioeconomic category, the split was
always between either equitable and equal access or inequitable and equal access. No
socioeconomic variables had results with opposing inequitable and equitable results
within the same access measure. This indicates that even in situations where variable
selection affects the results of the spatial or equity results, the affect is never so great that
it changes the result from one extreme to the other. However, that did occur between
access measures, meaning that a socioeconomic group might have equitable access by
one measure, but inequitable access by another.
More insight can be gained by looking at the overall consistency of the results
(see table 7). By not looking at whether results are significant or not, but only by if they
are consistent, insight is gained on variables’ effect on equity results. By combining the
results of all equity analyses into groups of four by the three access measures and six
socioeconomic variables, ten of the eighteen categories produced the same result
regardless of census unit and metric used (meaning all four combinations of variables
produced the same equitable, inequitable, or equal access result). Five of the remaining
78
eight categories had three similar results and one different. There was no apparent pattern
to which variable combination produced the inconsistent result. The last three category
groupings were split down the middle with two pairs of similar results. In these cases, all
three were split by census unit, not metric. Examining the results of the equity analyses
in this way helps illustrate what impact variable selection has on results. The majority of
the results are not affected by variable selection, but many of them are. Although this
does not indicate which methods are best for use in spatial and equity analyses, it does
indicate that variables should be selected carefully and with the understanding that they
could have an impact on results.
Another way of looking at result consistency is by comparing the variable
combinations to what I believe is the best option based on previous studies. For my indepth equity analysis I used blocks and the network metric because the most accurately
portray households and actual travel distance. However, network distance is more labor
intensive and requires a more expensive ArcGIS software package. Additionally, using
blocks is also more work because it affects the amount of data needing to be organized
and processed. If a case study were to implement only one of the other combinations due
to time or software restraints, they would still probably produce the same results the
majority of the time. In this study, when using block groups and Euclidean metric, by far
the easiest and least time-consuming variable combination, the same equity results were
produced as using blocks and network distance in over 60% of the socioeconomic
variable and access measure combinations (see table 6). Additionally, using the next
easiest combination, blocks and Euclidean produced the same equity results in over 80%
79
!
of the analyses, demonstrating that if future studies were to utilize a simpler method due
to time or software licensing restraints, the data would still be valuable.
Park Proximity
When measuring access through park proximity, it was found that although
variable selection did have an effect on the distance measured to the closest park, there
was little effect on the results of the equity analysis (see table 8 for spatial results, table 6
for equity results, and 9 for variable comparison results). When comparing the use of
blocks and block groups, the results of the Mann-Whitney U test showed that there was
an insignificant difference between the distributions of distance measurements between
the two census units (p=0.5). The median distance from block group centroid to park
boundary was 780 feet (N=958; SD=653). Similarly, the median distance from block
centroid was 740 feet (N=17046; SD=682). These spatial results imply that census unit
selection does not play a large role in the spatial measurement.
However, when comparing the use of Euclidean and network distances, results
indicated that there was a significant difference in the distributions of measurements
(p=0.00). Network distances were significantly higher with a median of 877 feet
compared to the Euclidean median of 640 feet (N=9002; Network SD=761; Euclidian
SD=563). The median difference between the two measurements for each census unit
was 189 feet (N=9002; SD=342). The median difference in distance for blocks is 191 feet
(N=8523; SD=345), while the difference for block groups is 162 feet (N=479;
SD=289.45). These results indicate that metric selection has a greater affect on spatial
results than census unit selection.
80
When combining the method variables to perform an equity analysis, it was found
that variable selection played little role in the significance of the results (See table 7 for
equity results). Five out of the six socioeconomic variables were unaffected, meaning that
in each of the four combinations of method variables, the results within each
socioeconomic variable were of the same significance level. Population density was the
only one with varying results. When measuring to the closest park the results were
different for analyses using blocks and block groups. Regardless of the metric used,
measurements using block groups were insignificant (p=0.7; p=1.0), whereas both using
blocks were significant (p=0.01; p=0.00). The level of significance being split by the
census unit used, and not by the metric used implies that for some variables, in this case
population density, census unit selection can affect the results, despite there not being a
significant difference between the two census units when analyzing spatial results.
Amount of Park Area
Results indicate that method variable selection has an effect on measuring the
amount of accessible park area (see table 8 for spatial results, table 6 for equity results,
and 9 for variable comparison results). Similar to park proximity, the results of the MannWhitney U tests show that the selection of census unit did not have a significant effect on
the amount of area measured (p=0.9). The median amount of area within a quarter mile of
block groups was 4.5 acres (N=960; SD=61). Blocks had a similar median of 3.7 acres
within a quarter mile (N=17046; SD=73). This implies that census unit has only a mild
effect on spatial results.
81
!
Results indicated that the distributions of the Euclidean and network metrics were
significantly different in regard to the amount of park area (p=0.00). The Euclidean
metric yielded significantly higher park area with a median amount of 9.5 acres (N=9003;
SD=89). Using the network metric produced a much smaller median of 0.8 acres within a
quarter mile (N=9003; SD=43). The median difference between Euclidean and network
measurements was 2.8 acres (N=9003; SD=72). The median difference in park area
between metrics for blocks is 3 acres (N=8523; SD=73), while the difference for block
groups is 0.3 acres (N=480; SD=45). These results imply that metric has a strong affect
on the amount of area measured. However, the fact that the difference between Euclidean
and network measurements for each census unit being that different, does seem to
indicate there is some difference, even if it did not return a significant value in the MannWhitney U test.
These differences in the spatial results by metric and census unit variable led to
mixed results in the four equity analyses that were run (See table 7 for equity results).
Percentage of renter occupied housing units and percentage of households with residents
under the age of 18 were the only socioeconomic categories that produced results of the
same significance despite variable selection (both were insignificant in all tests). The
other four socioeconomic categories produced mixed results, meaning that within each
socioeconomic category, the equity results using different variables did not produce
results of the same significance level (see table 10 for a visual of result consistency). The
results for each of socioeconomic categories were split with one variable combination
producing an equity result being of difference significance than the other three tests.
However, the odd one was not consistent from category to category. For example, When
82
looking at the amount of park area accessible to households with residents over 64, three
out of the four variable combinations produced results indicating that they have
significantly equitable access to park acreage. Only the equity result found using block
groups and the Euclidean metric returned an insignificant result indicating that
households with residents over 64 only have equal access to park area. In contrast to this,
when examining the results for population density, three out of the four variable
combinations produced results indicating that census units with high population density
have significantly inequitable access to park acreage. Only the equity result found using
blocks and the network metric produced an insignificant result indicating equal access.
These two examples both returned three matching results, and one different; however the
inconsistent result was found using different combinations of variables. This indicates
that variables affect the equity results of socioeconomic categories in different ways, with
no obvious pattern.
Park Service Area
Whether residents are considered inside or outside the quarter mile park service
areas seems to likewise be affected by the method variable selection (see table 8 for
spatial results, table 6 for equity results, and 9 for variable comparison results). Due to
the lack of unique measurement results in this access measurement, population density
was used to compare method variables. Following the same pattern as the other measures,
the results of the Mann-Whitney U tests indicate that the distributions of the residents
inside and outside the service area were not significantly different between blocks and
block groups (p>0.1). When comparing the percentage of census units that were inside or
83
!
outside the quarter mile threshold, blocks and block groups were within 3% margins of
each other within the metric being used (see table 10 for the number and percentage of
census units inside park service areas by metric). Similar to previous measures, results
indicate that the distributions of the data inside and outside the service area was
significantly different between Euclidean and network metrics (p=0.04). An additional
13% more blocks and block groups were measured to be within the service area when
using the Euclidean metric compared to network. These results indicate that metric
variable has a much greater affect on the spatial results than the census unit variable.
When paired with socioeconomic data for the equity analyses, the selection of
method variables had some affect on the significance of results. Three out of the six
socioeconomic variables tested maintained the same significance in each analysis despite
variable use. These were percent vacant housing, percent renter occupied housing, both
of which were considered significantly equitable throughout, and lastly percent
households with residents under the age of 18, which remained significantly inequitable
(p<0.02 for all). Percent non-white residents and percent households with residents over
64 were both split with two significant results and two insignificant. In both these
categories the significant results were in the two tests using blocks. Population density
was split with three significant results and one insignificant, which used block groups and
Euclidean metrics (see table 10 for a visual of result consistency). These results indicate
that despite metric having a significant influence on spatial results, it seems that census
unit had a greater effect on the equity results.
84
85
!
!
Table 6: Results of equity analyses using all combinations of census units, metrics, and access measures. Green indicates
equity with a significant p-value, red indicates inequity with a significant p-value, yellow indicates equality with an
insignificant p-value
86
Table 7: Results of equity analyses grouped by spatial analysis and socioeconomic variable to show consistency of results.
Each colored box represents the consistency of results for four Mann-Whitney U tests.!
87
!
Table 8: Descriptive statistic results of spatial analyses using all combinations of census units, metrics, and access measures.
Table 9: Results of method comparison analyses. Compares blocks with block groups,
and Euclidean with network metrics to determine if method selection plays a significant
role in spatial analyses. Bold indicates a significant p-value (p<0.05).
!
Table 10: The number and percentage of census units inside and outside of park service
areas when using Euclidean and network metrics.
88
CHAPTER 5: DISCUSSION
My overall aim of this study was to determine if traditionally disadvantaged
residents of Seattle live within an equitable distance to similar amounts of public park
acreage and whether the methods used to determine this affect the outcome of the results.
Through the use of spatial and statistical analyses I found that most of the traditionally
disadvantaged groups examined in this case study have equitable or equal access to parks,
but that method selection can play a role in the strength of results. Due to the two-part
nature of this study’s objective, I will discuss the results in two sections. First the equity
analysis using blocks and network metrics will be discussed, followed by a closer look at
the method variable analysis. Following these discussions I will go over some of the
limitations I experienced in this thesis, as well as my recommendations for future studies.
PARK ACCESS AND SOCIOECONOMIC EQUITY DISCUSSION
Access to parks appears to be equitable for the majority of the disadvantaged
groups studied. Using the three measures of access, and the six socioeconomic variables,
over half of the results using blocks and the network metric were equitable, and less than
a quarter were inequitable, with the remainder being of equal access. In this section I will
first discuss overall patterns and relationships followed by more in-depth examinations of
the three measures used to determine access.
89
!
Overall Patterns
This Seattle case study helps further the knowledge of how these socioeconomic
variables and access measures compare across case study locations. Unlike other studies,
age was the only factor that showed inequity. Households with individuals under the age
of 18 are farther away from parks and only have an equal amount of park area. Older
residents live an equal distance from parks, but in areas with more park area. Considering
previous case studies, it was a pleasant surprise to find that non-white residents had equal
or equitable access to parks in all three measures. This could mean that Seattle has less
segregated communities than Baltimore, Maryland (Boone et al., 2009), Macon, Georgia
and Pueblo, Colorado (Talen, 1997) which were all found to have inequitable park access
for non-white residents. Or perhaps simply there are just more parks in Seattle’s majority
non-white areas. A subsequent study could split the non-white category into individual
race and ethnicities to perform a follow-up equity analysis to see how the results differ by
more precise categories. Another explanation for the results could be that historically
non-white neighborhoods are becoming more gentrified and mixed, meaning that results
could be skewed causing less perceived inequity. Wolch et al. (2014) discuss the ‘green
paradox’ of how park creation in non-white communities can lead to the gentrification of
the neighborhood. This prompts a further study that could be completed in Seattle using
historical census and park data to see how spatial access and community integration has
changed over time.
By comparing results with those from the Seattle Parks and Recreation (SPR)
Legacy Plan survey (SPR, 2014), new insight is gained on how residents use parks and if
spatial access is a limiting factor or if perhaps there is another issue limiting park use.
90
The SPR survey results indicate that non-white residents were somewhat more likely than
white residents to visit a park more than once a week. This corresponds with the results
of the equity analysis showing that non-white residents have equal or equitable access to
parks. It cannot be assumed that just because there is better spatial access to parks for a
group, that they will visit a park more often. However, the fact that the survey showed
non-white residents visiting parks often and this study showing they have equal and
equitable park access demonstrates that at least spatial access is not a barrier of use, and
may be even promoting it.
Unfortunately, the survey found that lower-income residents are much less likely
to visit parks than high-income residents. The results of this study used the income proxy
variables of vacant housing and renter occupied housing which both resulted equitable or
equal access in all measures. As mentioned above, spatial access does not necessarily
translate to use, and the pairing of these results suggest that some other factor besides
spatial access, such as time and money, is limiting park use. Further study into what nonspatial factors are limiting access would help determine how park use could be promoted
for low-income residents.
As for older residents, the survey showed that adults over age 55 use parks less
than other adults. This raises the question of what limits their use of parks. The equity
analysis indicates that while households with older residents have more nearby park
acreage than other households, they only live an equal distance to parks and are more
often outside of the park service areas than other households. Although it is conjecture,
perhaps for older adults, the distance to the park is more limiting in terms of spatial
access, but more likely there are other non-spatial factors also impacting park use.
91
!
However since this study did not look into park use, further studies would need to be
conducted to understand what other factors to determine what could promote park use for
older residents.
Lastly, the survey indicates that households with children use parks at a much
higher rate than households without children. As discussed above, this study found
households with individuals under the age of 18 to live farther from parks and only have
an equal amount of area as households without individuals under 18 years of age. This
suggests that households with children are more willing to travel farther to visit parks.
Although it is encouraging that children are still potentially benefiting from visiting
parks, the disconnect between visitation rates and proximity shows this is an area that
SPR could attempt to improve. Or perhaps, because children are still gaining access to
public parks despite their poor spatial access, it allows SPR to focus on improving
visitation rates for other groups that are not visiting parks even with good spatial access,
such as low-income and older residents who are using parks less despite mostly equal and
equitable access.
By pairing the results of the visitation survey and this analysis of spatial access,
new understanding of the complexities of access comes to light. Other factors besides
spatial access affect visitation for Seattle residents. Future studies could focus on the
communities indicated here to help gain insight on what other factors are limiting park
use, and how much, if at all, the spatial relationship affects visitation.
Overall Seattle has fairly comparable results with previous access and equity
studies. Some communities have better access than others, but the majority has equal or
equitable access. Although pairing the results of this study with the SPR survey helps
92
clarify some aspects about the relationship between spatial access and park visitation, it
raises many more questions that could be looked at in the future.
Park Proximity
Using park proximity to determine access for residents in Seattle demonstrated
that although most residents are within a reasonable distance from their closest park, not
all socioeconomically disadvantaged residents have equitable access compared to their
privileged counterparts. The spatial analysis showed that even those households in the
quartile living the farthest from parks had a median under a half-mile. This indicates that
despite the fact that some disadvantaged groups live farther than their privileged
counterparts to parks, the majority are still within a reasonable distance. In fact, a halfmile is another threshold often used in access studies as an acceptable walking distance
(Nicholls, 2001; TPL, 2017).
The results of the Mann-Whitney U tests showed mixed results. While many
disadvantaged groups have equal or equitable access to park, the youth population was
found to have inequitable access. The result showing households with residents under 18
being significantly farther from parks was surprising. Most studies also looking at spatial
access to public parks found significantly better access (Nicholls & Shaffer, 2001), or
equal access for the youth population (Nicholls, 2001; Talen, 1997). Only Wolch et al.
2005 found children to have poor access in the city of Los Angeles, California. Seattle
appearing to having inequitable access for youths may be exaggerated by this study only
using public municipal parks. Access to schoolyards would help to fill the void of parks
93
!
in many ways. Including schools in a spatial analysis focused on Seattle youth might shed
light on the extent of inequity of access.
Amount of Park Area
Measuring park area demonstrated that most of Seattle has access to very little
park acreage, but that traditionally disadvantaged residents of Seattle do have equal or
equitable access compared to their privileged counterparts. However, looking at the
spatial analysis, it was unexpected to find that over half of the blocks examined had less
than an acre of park space within a quarter mile distance. Especially when taking into
consideration the result from the previous measure that the majority of Seattle residents
are less than a half-mile away from parks, this indicates that although there are nearby
parks, they are very small. Additionally, the high discrepancy between median (0.81
acres) and mean park acreage (15.3 acres), indicates that although there are large parks,
they are not within walking distance for most residents. Of course it cannot be assumed
that people only visit the park closest to their residence, especially if it is a small park
with less appealing features. It could be possible that residents may be more willing to
travel farther to visit large park destinations, so even if residents appear to have very little
park space near to where they live it does not mean that they are not visiting parks and
gaining benefits from those visits. However, it is important to have nearby park acreage
that does not require a large amount of planning, time, or money to visit, especially for
residents who might be limited by those factors.
It was a pleasant surprise that all disadvantaged groups in Seattle had equitable or
equal access to similar amounts of park acreage. This was somewhat unexpected since
94
other studies using park area as an indicator of access had mixed results. For example,
this case study found non-white residents of Seattle to have equitable access to park
acreage. In contrast, Boone et al. (2009) found that in Baltimore, access to park acreage
between race/ethnic groups was significantly different, with the white population having
access to a higher amount of acreage within a quarter mile than the African American
population. Additionally, in 1997, Talen completed a pair of case studies using park area
in Pueblo, Colorado, and Macon, Georgia that found that non-white residents had
significantly equitable access to park area in Macon, and inequitable access in Pueblo.
Similar this study, Talen also found areas with vacant housing (which she used to
represent low-income areas) to have equitable access to park area in Macon, but only
equal amount of area in Pueblo. Also similar to this study, Talen (1997) found that the
youth population had equal access to park acreage in both of the locations. These results
further demonstrate how results can vary between case studies. While both of these other
studies had results indicating inequitable amounts of parks area near certain groups, this
Seattle case study only had equal and equitable results.
Park Service Area
Using park service areas as an indicator of access showed that while the majority
of people live within the service area of parks, some residents of Seattle do not have
equal access to parks. The spatial analysis showed that about 70% of blocks were
considered to be inside the quarter mile service areas around parks (for the sake of
comparison, it can be noted that for Seattle, the percentages of blocks, block groups and
residents were all near 70%). This figure is on the higher end compared to studies using
95
!
service areas. Other studies using a quarter-mile threshold have found 26% of block
groups in Baltimore (Boone et al., 2009) and 29% of residents in Los Angeles were
inside park service areas (Wolch et al., 2005). Similar to Seattle, Talen found that 71% of
Chicago residences were within a quarter mile of parks (Talen, 2010). Nicholls (2001)
used a half-mile threshold but still only found that 38% of blocks were in the service
area. While these case studies use different census unit variables, the comparison still
indicates that Seattle is on the high end for the amount of residents inside park service
areas.
After the mix of significant and insignificant results produced by the two previous
measures in this study as well as the mixed service area results found in previous studies,
it was unanticipated for the equity analysis to produce significant results across all
socioeconomic variables in this measure. In Seattle, both the youth and elderly
populations appear to have inequitable access to parks. A study from Los Angeles is the
only other research using this variable and quarter-mile service areas that found children
to have poor access to parks (Wolch et al., 2005). Nicholls (2001) found the youth
population to have equitable access and older population to have equal access using a
half-mile threshold in Bryan, Texas. In her study with Shaffer, Nicholls (2001) found
both age groups to have equitable access to parks in College, Texas. This Seattle case
study found socioeconomically disadvantaged groups related to race, income (based on
vacancies and renter occupation), and population density to have equitable park access.
Interestingly, using similar methods, the studies conducted by Boone et al. (2009) in
Baltimore, Maryland, and Nicholls (2001) in Bryan, Texas found similar results with
comparable variables in their equity analyses.
96
METHOD VARIABLE ANALYSIS DISCUSSION
The second aim of this case study was to determine if the choice in variables
impact the outcome of the spatial and equity analyses. The results of this Seattle case
study show that method variable selection can play a role in the outputs of the spatial
analysis as well as the significance of results in the equity analysis. First I will discuss the
overall patterns found in this analysis, and then I will go more in-depth about the three
measures used to determine access.
Overall Patterns
For this methodology analysis I used methods from other studies, but combined
them in new ways to dig deeper into what variables affect spatial and equity analyses
results. Park proximity had the least variation, with five of the six socioeconomic
variables producing consistent results. Park service areas were next with three
unchanging categories. Amount of park area was the most variable, with only two
unchanging, and little pattern to the mixed results. This demonstrates that some measures
more than others are affected by variable selection. The use of more than one measure
can help highlight different aspects of spatial access. Boone et al. (2009) found in
Baltimore, Maryland that although African Americans had to travel less distance to parks
than white residents, that white residents had significantly more park area near them. In
this Seattle case study, I found that many of the socioeconomic categories had equity in
one measure but not another, depending on the method variables used. This highlights the
importance of using more than one measure to give a better picture of the access residents
have to parks.
97
!
Consistently the secondary variable comparison analyses indicated that census
unit selection did not affect the spatial measurement and that metric did (see table 9).
However the equity results point the other way, showing three categories that are split
into pairs by the census unit, and none by the metric used. Interestingly, all three of these
are also significant using blocks and insignificant using block groups. This perhaps
indicates that blocks, being the smaller aggregation level of households, highlight the
socioeconomic diversity in a different way than the larger aggregate block groups can.
Since no other studies compared origin point variables, it is difficult to know if this
pattern would exist outside of Seattle. Additional case studies performing similar
analyses would help further the knowledge from the baseline set by this study.
Overall, regardless of method, Seattle demonstrated equal and equitable access to
parks for traditionally disadvantaged groups. This methods analysis shows that method
selection does matter, but not in the same way for every measure or variable. Although
blocks and network distance are the best representation of households and travel distance,
using other variable combinations produced the same results 60% to 80% of the time.
This demonstrates that using more than one method is always the best practice since it
can test the strength of the results.
Park Proximity
When using park proximity to measure access, the results were by far the most
consistent regardless of variable combination (See Tables 6 and 7). Five out of six of the
socioeconomic variables had the same outcome, regardless of the census unit and metric
98
used. However the results for population density in the equity analysis demonstrates that
variable selection can matter.
The variable comparison analysis indicated that the distance measurements were
not significantly different between blocks and block groups, but unexpectedly, results for
population density were split by analyses using blocks and block groups. The results of
these analyses do not seem to match. The results show that census unit does not affect the
distances measured to parks, but it does affect population density. However, this logically
makes sense. Due to the small size of blocks compared to block groups, there are units
with very small populations and units with high populations, which creates extreme
outlier high and low densities. This affects the range of the data, which is what the MannWhitney U test looks at. However, to further understand what caused this split, the data
was further examined. When comparing the raw density data of block and block groups
the means are only off by thirty. However, comparing the medians reveals a difference of
over 550. Additionally, the maximum density for the block data is over four times as high
as the block group data and the minimum is over fifty times smaller. This helps explain
why the use of census unit is insignificant in the spatial analysis, yet still played a role in
the population density equity results.
As could be predicted, Euclidean distances were much shorter than network
distances. The result of the Mann-Whitney U test confirmed this, showing that the ranges
of distances measured were significantly different. This confirms the results from Sander,
Ghosh, van Riper, and Manson (2010), which also found the measurements to be
significantly different, with a median difference of 600 feet, much larger than the median
difference of less than 200 feet found in Seattle (Sander, Ghosh, van Riper, & Manson,
99
!
2010). Interestingly, this difference in distance measurement did not have an effect on the
results of the equity analyses. This was somewhat surprising considering the results from
the study by Higgs et al. (2012) that found significant results using Euclidean metrics, but
not network metrics in their comparative equity analyses.
Depending on what socioeconomic factors are used, it seems that when using park
proximity to measure access, the census unit and metric choices have little effect on
results. As Higgs et al. (2012) and other researchers in the field emphasize, it is important
to use varying methods to test results for strength. For the Seattle equity analysis
discussed above using blocks and network distance, the other three variable combinations
confirm and strengthen the results, with density being the only somewhat debatable
result. These results also indicate that if a future study were to have time or resource
constraints that a less labor-intensive method could be used to measure park proximity
with the assumption that results would be nearly the same in the equity analysis.
Amount of Park Area
Using the amount of park area within a quarter mile to measure access produced
the most inconsistent results (See Tables 6 and 7). Only two of the six socioeconomic
categories had the same outcome for each method variable combination. The other four
categories had three similar results and one different, which was not consistent to any
variable combination. These inconsistences demonstrate that method variable selection
can have a widespread affect on the results of equity analyses.
The secondary variable comparison analysis indicates that census unit selection
plays an insignificant role and metric selection plays a significant role in the amount of
100
park area calculated (see table 9). The raw data and medians support this, with census
unit medians that are less than one acre different, and metric medians that are nearly ten
acres different. Interestingly, the median difference between Euclidean and network
measurements was less than three acres. Even looking more closely at each combination
of variables does not reveal any reasoning behind why one result from four
socioeconomic categories differs from the others. The median distances range from less
than an acre to almost ten acres, and although it is interesting that there is so much
variation caused by method selection, it still does not shed light on a pattern in the equity
analysis results.
Unfortunately there are no other studies that utilize park area in a cross variable
equity analysis, so it is difficult to know if these inconsistent results are seemingly
random or to be expected. For example, Talen (1997) used network distances of one and
two miles from block centroids and found mixed results. Additionally, Boone et al.
(2009) used a Euclidean distance of a quarter mile from block group centroids and found
racial inequity in Baltimore. However these studies only use one census unit and metric
making them poor comparisons.
The mixed results found through combining variables in this study further
demonstrates how method standardization and utilization of multiple methods can bring
new understanding to the results of an equity analysis. Two of the conflicting results fall
in the analysis using blocks and network distance, which was used for the in-depth equity
analysis discussed above. Although these variables seem to make the most logical sense
due to their more accurate representation of people and travel distance, the fact that the
other three combinations point to a different result does raise some concern. However,
101
!
the fact that there are other conflicting results using other combinations, makes it seem
more random than not.
Park Service Area
When using park service areas to measure access, the results followed the same
pattern with some socioeconomic categories being more mixed than others (See tables 7
and 9). Three of the six categories produced consistent results regardless of method
variable combination. Two others were split into pairs of matching results, and the last
had three agreeing results and one conflicting. These results continue to demonstrate
how variable selection can affect the outcome of equity analyses.
The secondary census unit and metric analysis indicated that census unit selection
does not affect results (see table 9). Looking at the total number of census units included
in the service area supports this (see table 10). For block groups and blocks the
percentage of both units and population inside the service area were very close with all
between 75% and 78%. However, equity results seem to indicate that there is a difference
between using these census units. Similar to the division in the population density results
in park proximity, the conflicting pairs of results are separated by the analyses using
blocks and block groups. This occurred in the non-white residents and residents over 64
categories. Both had significant results in Mann-Whitney U tests using blocks, and
insignificant using block groups. By further examining the raw data to compare the
medians of the socioeconomic attributes in the four analyses, some insight can be gained.
The differences in medians for percentage of non-white residence between the blocks
inside and outside the service area were somewhat closer than that of block groups, but
102
the same cannot be said for percentage of residents over 64. However looking at the
distribution of data, which is what the Mann-Whitney U test uses, the blocks consistently
have a much larger range. This, along with the much larger sample size of blacks than
block groups, could be part of the reason for the division in results. Additionally, due to
the use of percentages instead of raw data in the analyses, blocks are more prone to
outliers. Blocks have smaller population sizes than block groups, so although the use of
percentages helps normalize the data for comparison, it makes blocks appear more
extreme then block groups. For example, it is much more likely to get a 100% value for a
socioeconomic characteristic using blocks than block groups if the population is small.
For each distribution in both of the split socioeconomic categories, blocks had a range of
100, whereas block groups ranged in the eighties for non-white residents and fifties and
sixties for households with residents over 64. This helps shed light on the division in the
results, showing that although blocks and block group selection may not affect some
categories and measures, it does others. Unfortunately there are no other studies that
utilize both of these census units, so there is no way to determine if these differences
would occur in other studies.
Results of the secondary method variable analysis indicated that there is a
significant difference between Euclidean and network service areas. The data for blocks
and block groups reflects this with around 84% of units being included when using
Euclidean, and only around 69% when using network distance, with an average
difference of around 14%. Nicholls found similar results, with nearly a 17% gap between
the amounts of block groups within a half-mile service area using both metrics (2001). In
her study with Shaffer, even more extreme results were returned, with a 25% gap
103
!
(Nicholls & Shaffer, 2001). However the equity analyses did not seem to be profoundly
affected by this difference. There was no pattern in the medians or ranges indicating that
the metric selection affected the significance of the results. This was unexpected based on
the results of previous studies. When using blocks and comparing the use of half mile
Euclidean and network service areas, Nicholls (2001) found that some socioeconomic
variables were affected. Population density and percent non-white residents were found
to have equitable access regardless of method. However percent of residents over 64 was
only equitable using Euclidean, and percent of residents under 18 and percent renter
occupied only when using network (Nicholls, 2001). These studies show how methods
may affect one case study differently than another. Nicholls (2001) concluded that
network was more accurate and better represents how people travel to parks. This
logically makes sense, and although the equity analysis does not indicate a significant
difference in the results, the amount of census units included in the service areas does.
LIMITATIONS AND RECOMMENDATIONS
As with any study, insight on the shortcomings and limitations of the study
became apparent throughout the process. These can be useful for future research and case
studies, so as to better understand methods or try to find a better way. Some issues were
mentioned in other studies, providing insight into this study’s own methodology.
One limitation of many access studies is border effect caused along the outer
perimeter of the study area. This study only used data from within the city limits. Only
Seattle census data, parks, and streets were used. Because Seattle is bordered by water to
its east and west sides, this really only affected the north and south boundaries. All of the
104
census units along the edges may have had closer parks or more park area within the
quarter mile distance threshold if park from adjacent communities had been included. In
this study, the block with the greatest distance to the nearest park had to travel just under
a mile and a half. This block is located in an area of south Seattle that does not have a lot
of parks. However, this is exacerbated due to the uneven nature of Seattle’s southern
border. The block is located on a peninsula of Seattle, surrounded by suburbs. This block
most likely has other parks nearby, but in other municipalities. Additionally, parks that
are near the edge of city limits may have non-Seattle residents within their service areas
that were not included in this study, which may skew some of the socioeconomic
numbers. This border-effect limitation is common to any spatial study that uses a strict
line in its study area. Higgs et al. (2012) mentions it, as do Talen and Anselin (1998).
Kaczynski et al. (2009) took the precaution of adding a total of 19 additional parks that
were within 800 meters of the boarder of a neighborhood that they studied in Waterloo,
Canada. This would have been much more complicated in this Seattle case study, due to
multiple municipalities bordering the city to the north and south.
Another main limitation of this study was the use of proxies for income. In an
ideal world, there would be household median income data available from the decennial
census at the block and block group level. However, to protect privacy this information is
not publically available at a small aggregation unit. While vacancy and renter occupation
rates are used often as indicators of income (Nicholls, 2001; Talen, 1997), they should
not be taken as a direct income comparison. In a study from Harvard University, it was
shown that the proportion of renters has increased in recent decades at all income levels
and that many choose to rent over home ownership (Joint Center for Housing Studies
105
!
(JCHS), 2013). Additionally, although vacancy rates can reflect an areas income level, it
can also be skewed by other factors. 2010, the year from which census data was used for
this study, was a peak for vacancies due to economic downturn, with rates lessening since
then. Vacancies can also be an indicator of new housing growth (JCHS, 2013). Seattle
has had a population boom in recent years (USCB, 2010a), so new housing not yet filled
or old housing waiting for demolition could skew vacancy rates (JCHS, 2013). Finding a
better way to represent income for future studies would strengthen that aspect of the
equity results.
Also related to variable limitations, I chose to group together all non-white
residents as well as have broad age categories. Census data is available for more specific
race and age categories, but for the sake of limiting the number of variables, I chose to
consolidate these categories. Follow-up studies could easily split these grouping apart to
see is if spatial access results are more meaningful.
Another limit of this study that should be mentioned was that only spatial access
was examined. Obviously there are many other factors that affect park access and use, so
this study is limited with the conclusions in can make about overall access. This is true
especially in regard to the comparisons I made to the results of the SPR survey (SPR,
2014). That survey had a fairly small sample size, and did not use exactly the same
socioeconomic groups as I did, therefore the comparisons are limited. Not only does
having spatial access not mean that you will visit the park, but also those without spatial
access can still visit parks. However, the fact that a park is nearby and available even if it
is not being visited has value, as demonstrated in the discussion of park benefits in the
literature review chapter.
106
Overall, very few issues came up during the course of this study. Many of the
issues discussed above had been mentioned in previous studies, so they were
acknowledged from the beginning. Future studies should take into account the issues
experienced in this study, but understand that there is not always a way to fix them, as is
the case with the unavailability of income data.
107
!
CHAPTER 6: CONCLUSION
The overall aims of this study were to determine if traditionally disadvantaged
residents of Seattle live within an equitable distance to similar amounts of public park
acreage and whether the methods used to determine this affect the outcome of the results.
The simple answer is that most traditionally disadvantaged residents in Seattle have equal
or equitable access to parks, with some exceptions, and that the methods used to measure
access can affect the results.
The research questions of this study were answered by completing a spatial
analysis using GIS techniques and statistical analyses. Three measures of access were
used: the distances from residents to the closest park, the amount of nearby park acreage,
and the service areas around parks. An equity analysis was completed using MannWhitney U tests with data from the spatial analysis to determine if traditionally
disadvantaged groups had equal, equitable or inequitable access compared to their
traditionally privileged counterparts. Low-income (represented by vacancy and renter
occupation rates), non-white residents, children, older residents, and densely populated
areas were all examined as disadvantaged groups that benefit from park access. The
spatial and equity analyses were completed four times for each access measure using a
combination of different census unit origin points (block and block group) and metrics
for measuring (Euclidean and network). Through the use of using multiple method
variables the results of the spatial and equity analyses were compared to determine the
variation of outcomes. Using these methods it was possible to determine the level of
spatial access residents have to parks and how methods affect the results.
108
The results indicated that most disadvantaged groups have equal or equitable
access to parks. When looking at the results of the three access measures using blocks
and network distance, which are the most accurate representation of residents and travel
distance, almost all of the socioeconomic variables demonstrated having equal or
equitable park access. Only age related variables, residents under 18 or over 64 returned
inequitable results, the young age range in two measures, and the older age range in one.
This highlights an area that could be focused on for improvement. Targeted park creation
or expansion would help increase spatial access for these groups. Additional studies
could be completed focusing on how the youth and older populations use parks, if spatial
access or some other aspect is a limiting factor, and if there are other sources of green
space or outdoor recreation available to them.
Results generated through the comparison of method variables demonstrated that
in this study, the choice of variables affected the equity analysis results less than half the
time. Regardless of the method variables used, the significance of the results remained
the same within each pairing of measure and socioeconomic variable the majority of the
time. The use of metric affected the spatial values calculated in GIS, but census units did
not. However, census units seemed to affect the outcome of the equity analysis more than
metric did. Reworking the data, adding more specific or different census categories, or
applying different statistical methods in a future study could help gain insight on the
causation behind these differences.
This study contributes to the field of study in several ways. At the local scale, this
study helps shed light on one aspect of the socioeconomic equity of park access for the
residents of Seattle. These results could potentially help the Seattle parks department set a
109
!
baseline for their park system to work towards better serving those communities
experiencing inequity. An in-depth GIS based access and equity analysis has never been
completed in Seattle, or anywhere in the Pacific Northwest. Seattle has an extensive,
well-established, and respected park system, so having an understanding of its spatial
accessibility sheds light on how even an excellent park system can improve. Additionally,
Seattle has, and will most likely continue to go through rapid growth and demographic
change. This study provides a snapshot in time of how accessible the parks were to the
population in 2010. It would be fascinating to repeat this study after the 2020 decennial
census to see how the socioeconomic equity of spatial access has changed.
On a broader scale, this study provides an additional data set to be examined and
compared against for future case studies. One challenge of this study was finding similar
studies with which to compare and contrast methods and results. Although a few studies
have compared metrics, no studies within the field of park access have compared the use
of blocks and block groups. Many studies use one method to perform an equity analysis,
with no explanation of the technical aspects of the method or justification for its use.
Using block groups and Euclidean metrics is technically easier and often used, but it is
not the most accurate representation of people and distance traveled. This study
demonstrates how choosing the easy or simple option can affect the outcome of the study.
Higgs et al. (2012) strongly suggests the use of multiple methods to strengthen results.
The authors express a frustration that there is no standardized way of measuring access,
making cross case study comparisons difficult. Even if future studies do not employ
comparative methods, it is important to understand and state how method selection could
potentially be affecting results.
110
With more and more of the earth’s population becoming urban each year (United
Nations, 2014), municipal parks are becoming even more depended upon as a way for
people to access nature. As cities expand and become denser, greenspaces are at risk of
development. However, due to a better understanding of the benefits of green spaces, the
importance of having parks and green spaces accessible to residents has become
apparent. Individuals, communities, and entire cities gain health, social, economic, and
environmental benefits from parks (Kabisch, Qureshi, & Haase, 2015). Unfortunately
research shows that often not all residents in an urban area have equal access to parks,
and therefore the benefits gained from them. Past case studies have used a variety of
methods to demonstrate the inequity of access to parks in urban areas (Boone et al., 2009;
Nicholls, 2001; Talen, 1998). This case study sought to determine if residents in Seattle
have equitable access to parks and to what extent the methodology affects the results. It
was found that most of Seattle’s traditionally disadvantaged groups examined in this
study have equal or equitable spatial access, and that although methodology choice can
affect the spatial measurement, it has a limited and irregular affect on the equitability.
111
!
WORKS CITED
Abercrombie, L. C., Sallis, J. F., Conway, T. L., Frank, L. D., Saelens, B. E., &
Chapman, J. E. (2008). Income and Racial Disparities in Access to Public Parks
and Private Recreation Facilities. American Journal of Preventive Medicine,
34(1), 9–15. https://doi.org/10.1016/j.amepre.2007.09.030
Bedimo-Rung, A. L., Mowen, A. J., & Cohen, D. A. (2005). The significance of parks to
physical activity and public health: a conceptual model. American Journal of
Preventive Medicine, 28(2 Supplement 2), 159–168.
https://doi.org/10.1016/j.amepre.2004.10.024
Bolstad, P. (2012). GIS fundamentals: a first text on geographic information systems (4.
ed). White Bear Lake, Minn: Eider Pr.
Bolund, P., & Hunhammar, S. (1999). Ecosystem services in urban areas. Ecological
Economics, 29(2), 293–301. https://doi.org/10.1016/S0921-8009(99)00013-0
Boone, C. G., Buckley, G. L., Grove, J. M., & Sister, C. (2009). Parks and People: An
Environmental Justice Inquiry in Baltimore, Maryland. Annals of the Association
of American Geographers, 99(4), 767–787.
Bowler, D. E., Buyung-Ali, L., Knight, T. M., & Pullin, A. S. (2010). Urban greening to
cool towns and cities: A systematic review of the empirical evidence. Landscape
and Urban Planning, 97(3), 147–155.
https://doi.org/10.1016/j.landurbplan.2010.05.006
112
Cavanagh, J. E., Zawar-Reza, P., & Wilson, J. G. (2009). Spatial attenuation of ambient
particulate matter air pollution within an urbanised native forest patch. Urban
Forestry & Urban Greening, 8(1), 21–30.
https://doi.org/10.1016/j.ufug.2008.10.002
Centers for Disease Control and Prevention. (2001). Increasing physical activity: a report
on recommendations of the Task Force on Community Preventive Services
(MMWR No. Vol. 50, No. RR-18). Atlanta, GA: U.S. Department of Health and
Human Services. Retrieved from http://www.cdc.gov/MMWR/PDF/rr/rr5018.pdf
Chang, C., Li, M., & Chang, S. (2007). A preliminary study on the local cool-island
intensity of Taipei city parks. Landscape and Urban Planning, 80(4), 386–395.
https://doi.org/10.1016/j.landurbplan.2006.09.005
Chapin, F. S., Matson, P. A., & Mooney, H. A. (2002). Principles of terrestrial ecosystem
ecology. New York: Springer.
Chiesura, A. (2004). The role of urban parks for the sustainable city. Landscape and
Urban Planning, 68(1), 129–138.
https://doi.org/10.1016/j.landurbplan.2003.08.003
City of Seattle. (2017). City of Seattle Open Data portal. Retrieved from
https://data.seattle.gov/
City of Seattle, & Department of Planning and Development. (2011). 2010 Census
Profile: City of Seattle. Retrieved from http://www.seattle.gov/
113
!
Cohen, A. J., Anderson, H. R., Ostro, B., Pandey, K. D., Krzyzanowski, M., Künzli, N.,
… others. (2004). Urban air pollution. Comparative Quantification of Health
Risks, 2, 1353–1433.
Cohen, P., Potchter, O., & Schnell, I. (2014). A methodological approach to the
environmental quantitative assessment of urban parks. Applied Geography, 48,
87–101. https://doi.org/10.1016/j.apgeog.2014.01.006
Comber, A., Brunsdon, C., & Green, E. (2008). Using a GIS-based network analysis to
determine urban greenspace accessibility for different ethnic and religious groups.
Landscape and Urban Planning, 86(1), 103–114.
https://doi.org/10.1016/j.landurbplan.2008.01.002
Corder, G. W., & Foreman, D. I. (2014). Nonparametric Statistics A Step-by-Step
Approach (2nd ed.). Hoboken: Wiley.
Crompton, J. L. (2005). The impact of parks on property values: empirical evidence from
the past two decades in the United States. Managing Leisure, 10(4), 203–218.
https://doi.org/10.1080/13606710500348060
Cutts, B. B., Darby, K. J., Boone, C. G., & Brewis, A. (2009). City structure, obesity, and
environmental justice: An integrated analysis of physical and social barriers to
walkable streets and park access. Social Science & Medicine, 69(9), 1314–1322.
https://doi.org/10.1016/j.socscimed.2009.08.020
De Ridder, K., Adamec, V., Bañuelos, A., Bruse, M., Bürger, M., Damsgaard, O., …
Weber, C. (2004). An integrated methodology to assess the benefits of urban
114
green space. Science of The Total Environment, 334–335, 489–497.
https://doi.org/10.1016/j.scitotenv.2004.04.054
Dooling, S., Simon, G., & Yocom, K. (2006). Place-based urban ecology: A century of
park planning in Seattle. Urban Ecosystems, 9(4), 299–321.
https://doi.org/10.1007/s11252-006-0008-1
Gómez-Baggethun, E., Gren, Å., Barton, D. N., Langemeyer, J., McPhearson, T.,
O’Farrell, P., … Kremer, P. (2013). Chapter 11: Urban Ecosystem Services.
Urbanization, Biodiversity and Ecosystem Services: Challenges and
Opportunities (pp. 175–251). Springer Netherlands. Retrieved from
http://link.springer.com/chapter/10.1007/978-94-007-7088-1_11
González-Oreja, J. A., Bonache-Regidor, C., & de la Fuente-Díaz-Ordaz, A. A. D. la.
(2010). Far from the noisy World? Modeling the relationships between park size,
tree cover and noise levels in urban green spaces of the city of Puebla, Mexico.
Interciencia, 35(7), 486–492.
Harnik, P., & Simms, J. (2004). Parks: How Far Is Too Far? Planning, 70(11), 8–11.
Hartig, T., Mang, M., & Evans, G. W. (1991). Restorative Effects of Natural
Environment Experiences. Environment and Behavior, 23(1), 3–26.
https://doi.org/10.1177/0013916591231001
Higgs, G., Fry, R., & Langford, M. (2012). Investigating the implications of using
alternative GIS-based techniques to measure accessibility to green space.
115
!
Environment and Planning B: Planning and Design, 39(2), 326 – 343.
https://doi.org/10.1068/b37130
Jim, C. Y., & Chen, W. Y. (2009). Ecosystem services and valuation of urban forests in
China. Cities, 26(4), 187–194. https://doi.org/10.1016/j.cities.2009.03.003
Joint Center for Housing Studies. (2013). America’s Rental Housing: Evolving Markets
and Needs. Harvard University. Retrieved from
http://www.jchs.harvard.edu/research/publications/americas-rental-housingevolving-markets-and-needs
Kabisch, N., & Haase, D. (2014). Green justice or just green? Provision of urban green
spaces in Berlin, Germany. Landscape and Urban Planning, 122, 129–139.
https://doi.org/10.1016/j.landurbplan.2013.11.016
Kabisch, N., Qureshi, S., & Haase, D. (2015). Human–environment interactions in urban
green spaces: A systematic review of contemporary issues and prospects for future
research. Environmental Impact Assessment Review, 50, 25–34.
https://doi.org/10.1016/j.eiar.2014.08.007
Kaczynski, A. T., Potwarka, L. R., & Saelens, B. E. (2008). Association of Park Size,
Distance, and Features With Physical Activity in Neighborhood Parks. American
Journal of Public Health, 98(8), 1451–1456.
https://doi.org/10.2105/AJPH.2007.129064
Kaczynski, A. T., Potwarka, L. R., Smale, B. J. A., & Havitz, M. E. (2009). Association
of Parkland Proximity with Neighborhood and Park-based Physical Activity:
116
Variations by Gender and Age. Leisure Sciences, 31(2), 174–191.
https://doi.org/10.1080/01490400802686045
Kaplan, R. (1985). The analysis of perception via preference: a strategy for studying how
the environment is experienced. Landscape Planning, 12(2), 161–176.
Kaźmierczak, A. (2013). The contribution of local parks to neighbourhood social ties.
Landscape and Urban Planning, 109(1), 31–44.
https://doi.org/10.1016/j.landurbplan.2012.05.007
Kaźmierczak, A., & Cavan, G. (2011). Surface water flooding risk to urban communities:
Analysis of vulnerability, hazard and exposure. Landscape and Urban Planning,
103(2), 185–197. https://doi.org/10.1016/j.landurbplan.2011.07.008
King County. (2017). King County GIS Data portal. Retrieved October 9, 2017, from
http://www5.kingcounty.gov/gisdataportal/
Klingle, M. W. (2007). Emerald city: an environmental history of Seattle. New Haven:
Yale University Press.
Konijnendijk, C. C., Annerstedt, M., Nielsen, A. B., & Maruthaveeran, S. (2013).
Benefits of urban parks: a systematic review. A report for IPFRA. IFPRA.
Koohsari, M. J., Kaczynski, A. T., Giles-Corti, B., & Karakiewicz, J. A. (2013). Effects
of access to public open spaces on walking: Is proximity enough? Landscape and
Urban Planning, 117, 92–99. https://doi.org/10.1016/j.landurbplan.2013.04.020
117
!
Kruckeberg, A. R. (1998). The natural history of Puget Sound country. Seattle:
University of Washington Press.
Kuittinen, M., Moinel, C., & Adalgeirsdottir, K. (2016). Carbon sequestration through
urban ecosystem services: A case study from Finland. Science of The Total
Environment, 563–564, 623–632. https://doi.org/10.1016/j.scitotenv.2016.03.168
La Rosa, D. (2014). Accessibility to greenspaces: GIS based indicators for sustainable
planning in a dense urban context. Ecological Indicators, 42, 122–134.
https://doi.org/10.1016/j.ecolind.2013.11.011
Leslie, E., Coffee, N., Frank, L., Owen, N., Bauman, A., & Hugo, G. (2007). Walkability
of local communities: Using geographic information systems to objectively assess
relevant environmental attributes. Health & Place, 13(1), 111–122.
https://doi.org/10.1016/j.healthplace.2005.11.001
Maller, C. J., Townsend, M., St Leger, L., Henderson-Wilson, C., Pryor, A., Prosser, L.,
& Moore, M. (2008). Healthy Parks Healthy People: The Health Benefits of
Contact with Nature in a Park Context - A Review of Current Literature (2nd ed.).
Burwood, Melbourne: Faculty of Health, Medicine, Nursing, and Behavioural
Sciences, Deakin University.
Markevych, I., Tiesler, C. M. T., Fuertes, E., Romanos, M., Dadvand, P.,
Nieuwenhuijsen, M. J., … Heinrich, J. (2014). Access to urban green spaces and
behavioural problems in children: Results from the GINIplus and LISAplus
118
studies. Environment International, 71, 29–35.
https://doi.org/10.1016/j.envint.2014.06.002
National Recreation and Park Association. (2017). 2017 NRPA Agency Performance
Review: Park and Recreation Agency Performance Benchmarks. Ashburn, VA:
National Recreation and Park Association. Retrieved from
https://www.nrpa.org/siteassets/nrpa-agency-performance-review.pdf
Nicholls, S. (2001). Measuring the accessibility and equity of public parks: a case study
using GIS. Managing Leisure, 6(4), 201–219.
https://doi.org/10.1080/13606710110084651
Nicholls, S., & Crompton, J. L. (2005). The Impact of Greenways on Property Values:
Evidence from Austin, Texas. Journal of Leisure Research, 37(3).
Nicholls, S., & Shafer, C. S. (2001). Measuring accessibility and equity in a local park
system: the utility of geospatial technologies to park and recreation professionals.
Journal of Park and Recreation Administration, 19(4), 102–124.
Nowak, D. J. (2002). The Effects of Urban Trees on Air Quality. USDA Forest Service.
Retrieved from http://www.nrs.fs.fed.us/units/urban/localresources/downloads/Tree_Air_Qual.pdf
Nowak, D. J., Greenfield, E. J., Hoehn, R. E., & Lapoint, E. (2013). Carbon storage and
sequestration by trees in urban and community areas of the United States.
Environmental Pollution, 178, 229–236.
https://doi.org/10.1016/j.envpol.2013.03.019
119
!
Nutsford, D., Pearson, A. L., & Kingham, S. (2013). An ecological study investigating
the association between access to urban green space and mental health. Public
Health, 127(11), 1005–1011. https://doi.org/10.1016/j.puhe.2013.08.016
Office of Financial Management, & Forecasting & Research Division. (2016). State of
Washington 2016 Population Trends. Washington State Office of Financial
Management. Retrieved from http://www.ofm.wa.gov/pop/april1/poptrends.pdf
Oh, K., & Jeong, S. (2007). Assessing the spatial distribution of urban parks using GIS.
Landscape and Urban Planning, 82(1–2), 25–32.
https://doi.org/10.1016/j.landurbplan.2007.01.014
Sander, H. A., Ghosh, D., van Riper, D., & Manson, S. M. (2010). How do you measure
distance in spatial models? An example using open-space valuation. Environment
and Planning B: Planning and Design, 37(5), 874–894.
Seattle Board of Park Commissioners. (1909). Seattle Municipal Archives Map Index.
Retrieved from http://clerk.ci.seattle.wa.us
Seattle Parks and Recreation. (2014). Parks Legacy Plan: Goals and Strategies. Retrieved
from
http://www.seattle.gov/Documents/Departments/ParksAndRecreation/PoliciesPla
nning/PlanningParksLegacyPlan.pdf
Sherer, P. M. (2003). The Benefits of Parks: Why America Needs More City Parks and
Open Space (White Paper). The Trust For Public Land. Retrieved from
https://www.tpl.org/benefits-parks-white-paper
120
Talen, E. (1997). The Social Equity of Urban Service Distribution: An Exploration of
Park Access in Pueblo, Colorado, and Macon, Georgia. Urban Geography, 18(6),
521–541. https://doi.org/10.2747/0272-3638.18.6.521
Talen, E. (1998). Visualizing Fairness: Equity Maps for Planners. Journal of the
American Planning Association, 64(1), 22–38.
https://doi.org/10.1080/01944369808975954
Talen, E. (2010). The Spatial Logic of Parks. Journal of Urban Design, 15(4), 473–491.
https://doi.org/10.1080/13574809.2010.502335
Talen, E., & Anselin, L. (1998). Assessing spatial equity: an evaluation of measures of
accessibility to public playgrounds. Environment and Planning A, 30(4), 595–
613.
Taylor, A. F., & Kuo, F. E. (2009). Children With Attention Deficits Concentrate Better
After Walk in the Park. Journal of Attention Disorders, 12(5), 402–409.
https://doi.org/10.1177/1087054708323000
Thrush, C. (2007). Native Seattle: histories from the crossing-over place. Seattle:
University of Washington Press.
Trust For Public Land. (2011). The Economic Benefits of Seattle’s Park and Recreation
System. Retrieved from
https://www.recpro.org/assets/Library/Economic_Impact/economic_benefits_seatt
les_parks-rec_system_2011.pdf
121
!
Trust For Public Land. (2016). 2016 City Park Facts. The Trust For Public Land.
Retrieved from https://www.tpl.org/2016-city-park-facts
Trust For Public Land. (2017). ParkScore Methodology. Retrieved March 31, 2017, from
http://parkscore.tpl.org/methodology.php
Ulrich, R. (1984). View through a window may influence recovery. Science, 224(4647),
224–225.
Ulrich, R. S. (1981). Natural Versus Urban Scenes Some Psychophysiological Effects.
Environment and Behavior, 13(5), 523–556.
https://doi.org/10.1177/0013916581135001
United Nations, Department of Economic and Social Affairs, & Population Division.
(2014). World urbanization prospects: the 2014 revision.
United States Census Bureau. (1994). Geographic areas reference manual. Washington,
D.C.: U.S. Dept. of Commerce, Economics and Statistics Administration, Bureau
of the Census. Retrieved from https://www.census.gov/geo/reference/garm.html
United States Census Bureau. (2010a). 2010 American Community Survey 1-Year
Estimates. Retrieved from https://factfinder.census.gov
United States Census Bureau. (2010b). 2010 United States Census. Retrieved from
https://factfinder.census.gov
United States Census Bureau. (2014). 2014 American Community Survey 1-Year
Estimates. Retrieved from https://factfinder.census.gov
122
United States Census Bureau. (2015). 2015 American Community Survey 1-Year
Estimates. Retrieved from https://factfinder.census.gov
Velasco, E., Roth, M., Norford, L., & Molina, L. T. (2016). Does urban vegetation
enhance carbon sequestration? Landscape and Urban Planning, 148, 99–107.
https://doi.org/10.1016/j.landurbplan.2015.12.003
Wang, D., Brown, G., & Liu, Y. (2015). The physical and non-physical factors that
influence perceived access to urban parks. Landscape and Urban Planning, 133,
53–66. https://doi.org/10.1016/j.landurbplan.2014.09.007
Williams, J. D. (2011). The 2010 Decennial Census: Background and Issues. No. R40551.
Congressional Research Service. Retrieved from
https://www.census.gov/history/pdf/2010-background-crs.pdf
Wolch, J. R., Byrne, J., & Newell, J. P. (2014). Urban green space, public health, and
environmental justice: The challenge of making cities “just green enough.”
Landscape and Urban Planning, 125, 234–244.
https://doi.org/10.1016/j.landurbplan.2014.01.017
Wolch, J., Wilson, J. P., & Fehrenbach, J. (2005). Parks and park funding in Los
Angeles: An equity-mapping analysis. Urban Geography, 26(1), 4–35.
Yang, L., Zhang, L., Li, Y., & Wu, S. (2015). Water-related ecosystem services provided
by urban green space: A case study in Yixing City (China). Landscape and Urban
Planning, 136, 40–51. https://doi.org/10.1016/j.landurbplan.2014.11.016
123
!
Yin, S., Shen, Z., Zhou, P., Zou, X., Che, S., & Wang, W. (2011). Quantifying air
pollution attenuation within urban parks: An experimental approach in Shanghai,
China. Environmental Pollution, 159(8–9), 2155–2163.
https://doi.org/10.1016/j.envpol.2011.03.009
Zhang, B., Xie, G., Zhang, C., & Zhang, J. (2012). The economic benefits of rainwaterrunoff reduction by urban green spaces: A case study in Beijing, China. Journal
of Environmental Management, 100, 65–71.
https://doi.org/10.1016/j.jenvman.2012.01.015
124
APPENDIX
Map of Seattle showing the percentage of vacant households per census block.
Results are split into quartiles.
125
!
Map of Seattle showing the percentage of renter occupied households per census block.
Results are split into quartiles.!
!
126
!
Map of Seattle showing the percentage of non-white residents per census block.
Results are split into quartiles.!
127
!
Map of Seattle showing the percentage of households with residents under the age of
18 per census block. Results are split into quartiles.!
!
128
!
Map of Seattle showing the percentage of households with residents over the age of 64
per census block. Results are split into quartiles.!
129
!
! Map of Seattle showing population density (residents per square mile). Results are split
into quartiles. !
130