Crime Under the Canopy: The Correlation Between Urban Forests and Crime Occurrence in Olympia, Washington

Item

Title

Eng Crime Under the Canopy: The Correlation Between Urban Forests and Crime Occurrence in Olympia, Washington

Date

2016

Creator

Eng Loosle, Wendy

Subject

Eng Environmental Studies

extracted text

CRIME UNDER THE CANOPY:
THE CORRELATION BETWEEN URBAN FORESTS
AND CRIME OCCURRENCE IN OLYMPIA, WASHINGTON

by
Wendy Loosle

A Thesis
submitted in partial fulfillment
of the requirements for the degree
Master of Environmental Studies
The Evergreen State College
June 2016

© 2016 by Wendy Loosle. All rights reserved.

This Thesis for the Master of Environmental Studies Degree
by
Wendy Loosle

has been approved for
The Evergreen State College
by

________________________
Kathleen M. Saul, MA, MES
Member of the Faculty

________________________
Date

ABSTRACT
Crime Under the Canopy: The Correlation Between Urban Forests
and Crime Occurrence in Olympia, Washington
Wendy Loosle
As urban populations continue to increase, the corresponding reduction in tree canopy
cover demonstrates notable impacts to community and environmental health. To address
the need for balance between nature and urbanization, research continues to seek ways to
incorporate nature within cities and explore the benefits of robust urban forests. An
increasing number of studies began recognizing that urban forests may provide the
additional benefit of crime suppression. However, a relatively small number of similar
studies have been conducted, and they are geographically localized or temporally limited.
Further research is needed to examine this relationship and assess the application in
spatial and longitudinal studies.
This research used data from the City of Olympia to focus on validating the relationship
between tree canopy cover and crime occurrence in an urban core area. Analysis was
conducted using regression, spatial, and temporal analysis at the census block group level
with socioeconomic variables as controls. Regression results from two annual datasets
demonstrated a significant inverse relationship, where a greater percentage of urban tree
canopy cover predicted reduced crime rates in 2008 (β = -13.114, p < 0.001) and in 2014
(β = -7.910, p < 0.001). This study concludes that incorporating high-canopied trees in
urban green space is associated with crime suppression. The data supports developing
cities to include urban forests and public green spaces as a way to reduce crime and
create healthier, happier communities.

Table of Contents
List of Figures ................................................................................................................. viii
List of Tables .................................................................................................................. viii
Acknowledgements .......................................................................................................... ix
Chapter 1: Introduction ................................................................................................... 1
1.1

Introduction .......................................................................................................... 1

1.2

Background .......................................................................................................... 2

Chapter 2: Literature Review .......................................................................................... 6
2.1

Introduction .......................................................................................................... 6

2.2

Physiological and Psychological Benefits of Exposure to the Natural
Environment ......................................................................................................... 6

2.2.1

Discussion of Theory .................................................................................... 6

2.2.2

Theoretical Synthesis and Application ......................................................... 8

2.3

Criminal Behavior and Motivations in Urban Areas ........................................... 9

2.3.1

Discussion of Theory .................................................................................... 9

2.3.2

Factors in Urban Crime ............................................................................... 11

2.4

Synthesis: Known Studies Linking Criminal Behavior and Mental States
Affected by Nature ............................................................................................. 12

2.4.1

Urban Green Space and Crime: The Conflict ............................................. 12

2.4.2

Urban Green Space, Mental Fatigue and Crime: Synthesis and the
Importance of Trees.................................................................................... 13

2.5

Conclusion: Gaps in Research and Advancement of the Field .......................... 16

Chapter 3: Methods ........................................................................................................ 17
3.1

Aim and Objectives of Research ........................................................................ 17

3.2

Study Site Description ........................................................................................ 17

3.3

Data .................................................................................................................... 20

3.3.1

Variables ..................................................................................................... 20

3.3.2

Data Preparation.......................................................................................... 24

3.4

Limitations ......................................................................................................... 27

3.4.1

LiDAR data ................................................................................................. 27

3.4.2

Variable Boundaries.................................................................................... 28

3.4.3

Crime Coding Effects ................................................................................. 29
vi

3.5

Statistical Analysis ............................................................................................. 30

Chapter 4: Results........................................................................................................... 33
4.1 Descriptive Statistics ............................................................................................... 33
4.2 Statistical Analysis Results ..................................................................................... 35
4.3 Discussion ............................................................................................................... 39
Chapter 5: Conclusion .................................................................................................... 42
5.1 Research Conclusions ............................................................................................. 42
5.2 Recommendations for Further Research ................................................................. 42
Works Cited ..................................................................................................................... 44
Appendix A ...................................................................................................................... 48
Appendix B ...................................................................................................................... 49
Appendix C ...................................................................................................................... 53

vii

List of Figures
Figure 1: Study area based on City of Olympia Police Department patrol sector “B” ......19
Figure 2: Census block group boundaries overlaying the downtown core of Olympia.....23
Figure 3: Tree canopy layer created using ArcMap 10.3 from 2008 orthoimagery ..........25
Figure 4: Tree canopy layer created using ArcMap 10.3 from 2015 orthoimagery ..........26
Figure 5: Choropleth map of 2008 block groups shaded by percent canopy cover ...........35
Figure 6: Geographically symbolized map of residuals for 2008 analysis ........................38
Figure 7: Geographically symbolized map of residuals for 2014-15 analysis...................39
Figure 8: City of Olympia Downtown Strategy Area ........................................................48
Figure 9: 2008 LiDAR Bare Earth return ..........................................................................49
Figure 10: 2008 LiDAR Highest Hit return .......................................................................50
Figure 11: 2008 high-intensity orthoimagery ....................................................................51
Figure 12: 2015 high-intensity orthoimagery ....................................................................52
Figure 13: Map demonstrating study limitations ...............................................................53

List of Tables
Table 1: Descriptive statistics for study area block groups in Olympia, Washington .......34
Table 2: Results from Ordinary Least Squares regression models ....................................36

viii

Acknowledgements
The research and analysis conducted as part of this thesis drew upon the generous
contributions of my colleagues, the Evergreen State College MES faculty, fellow MES
students, and several dedicated professionals.
I would especially like to thank my thesis adviser, Kathleen. Your boundless
encouragement, thoughtful insights, and patient ear ensured the success of my work.
Thank you for being a true mentor, cheerleader, and my friend.
No student is an island unto herself. My fellow students and friends in the MES cohorts
during the past three years inspired me personally and professionally. Thank you for your
friendship and support: Chelsea, Mehran, Sierra, Pam, Tiffany, Danae, Hannah, Michelle,
Daron, and all those who contributed feedback or a smile of encouragement.
The initial support for this research came from the City of Olympia’s urban forester,
Michelle Bentley. Thank you for providing me with continued encouragement and
networking with so many resourceful professionals. The data required to conduct this
research came from one such dedicated City of Olympia colleague, David Kuhn. Thank
you for supplying the “bones” of this thesis and always being available to answer
questions and connect me with other GIS-enthusiasts.
Considerable gratitude is due to Mike Ruth for inspiring my passion for GIS. Thank you
for your enthusiasm and for teaching me that there’s always a map in that.

ix

Chapter 1: Introduction
1.1 Introduction
As of 2016, the majority of the human population resides in cities; in America
alone, the urban population comprises 80% of all residents (U.N. Population Division,
2016). Not surprisingly, urban areas are expected to increase substantially over the next
50 years. Cities grow to accommodate new arrivals, both in terms of geographic
boundaries and in the ways that urban space is utilized to provide residential housing. As
a result, urban areas have lost more than 600 million trees to development over the last 30
years, demonstrating a 30% gross decline in tree canopy cover (American Forests, 2015).
To address the need for balance between nature and urbanization, practitioners
and scientists continue to seek ways to conserve and create green spaces within cities.
Drawing upon research that demonstrates the strong connection between mental health
and nature experience, inquiries into urban green space and its impacts on mental health
continue to provide evidence in support of incorporating nature as part of metropolitan
design. An increasing number of studies began recognizing that urban forests may
provide an additional, lesser-known benefit: crime suppression.
Previous studies examined urban tree canopy cover and crime occurrence at
varying geographic scales and, correspondingly, with varying methodology. The results
hold the potential to provide cities with beneficial and much needed information to garner
funding and public support; however, the relatively small number of studies are
geographically localized or temporally limited. Further research is needed to examine and
assess this potential relationship in spatial studies. Additionally, the analysis techniques
need further validation and no studies have been conducted that assess change over time.
1

The research conducted as part of this thesis seeks to answer the following
questions: Is there a relationship between the presence of urban forests and street trees
and the occurrence of crime in the City of Olympia? If a relationship exists, has it
changed over time? Building off of the established methodology from previous studies
and addressing the gaps in spatial and temporal analysis, this analysis addresses the
hypothesis that a greater percentage of tree canopy cover in an urban core area has an
inverse effect on crime, leading to reduced instances. Secondarily, while the study
focuses more on a repeated cross-section of time rather than a long-term time series, there
is value in understanding whether research conducted in one area is consistent over time.
This may uncover other potential confounding variables, and will speak to the strength of
the canopy/crime relationship.
Ultimately, this research aims to positively impact human and environmental
health in three ways by addressing the potential crime-reducing benefits of street trees: 1)
reducing stress and improving mental states before criminal behavior is induced, 2)
improving general feelings of security and well-being as a result of reduced crime, and 3)
support building better cities with green spaces as a way to reduce crime and create
healthier, happier communities.

1.2 Background
Romanticizing nature is thought to be the result of the benefits that cities provide
(Cronon, 1995; Jacobs, 1961), where the freedom from working the land and ease of
access to resources allowed people to dream about nature in a new way. The yearning to
be away from the “hustle and bustle” of the city led many to seek refuge in the interim

2

suburban areas, not too far from the cities resources but far enough to reach quietude. The
potential negative result, however, has been the suburban growth, prevalent in and since
the 1950s. This process involves destruction of the very aspect city-escapists were hoping
to find in “natural” settings.
Over the last several decades, however, the preference of our increasing
population shifted to residing primarily within metropolitan areas. In the last fifteen
years, urban populations rose from 46.6% to 54% (of the total world population), and by
2030 this number will likely range near 60% (U.N. Population Division, 2016). This
means that cities constitute the major living centers of the world, where humans primarily
find social and cultural interaction, where we find joy and sadness, where we live and die.
The process of urbanization involves converting green spaces to impermeable
surfaces, such as concrete and asphalt, removing trees and vegetation to make way for
human-built infrastructure. As defined by Fields (2002), densely forested urban areas
include at least fifty percent tree cover, whereas sparsely forested areas include less than
twenty percent tree cover. The urban forest cover of a typical American city in 2002 was
approximately thirty percent (Fields). Today, with broader participation in national
programs, such as the Arbor Day Tree City USA program, more investment into street
trees and maintenance is formalized through participants’ respective urban forestry
programs. Tools to quantify, manage, and plan urban forests have developed over the past
decade, evolving to online, GIS-based systems. Urban planners can easily access
evidence-based methods to develop multi-faceted design approaches that include nature
as an essential element (American Planners Association’s (APA) Green Community
Center, 2015).

3

Maintaining a high percentage of urban forest can positively affect a city in
several, well-documented ways: stormwater attenuation (Kirnbauer et al., 2013),
reductions in air pollutant concentrations (Fields, 2002; McPherson et al., 2005;
Kirnbauer et al., 2013), energy savings through mitigation of urban heat-island effects
(McPherson et al., 2005; Sydnor et al., 2011), reducing erosion, providing wildlife
habitat, and removing carbon dioxide (CO2) (McPherson et al., 2005; Kirnbauer et al.,
2013), CO2 sequestration (Livesley et al., 2014), and releasing oxygen into the
atmosphere (McPherson et al., 1999; Kirnbauer et al., 2013).
Additionally, substantial research demonstrates the positive impacts on human
health and behavior from “green spaces”, which may include urban vegetation, trees, and
other nature-based infrastructure or installations (Choudhry et al, 2015; Louv, 2012;
Velarde et al., 2007). Strong relationships with and continued connections with nature
support positive human growth, development, and sustainable human health. Severed
connections with nature, typically coupled with an overexposure to modern technology,
lead to unbalanced physical and emotional states, e.g., stress, fatigue, inability to focus or
complete tasks, and difficulty recalling information (Lohr & Pearson-Simms, 2006;
Louv, 2012; Velarde et al., 2007).
To add to this array of benefits, emerging literature assesses the relationship
between the presence of nature, specifically trees, and the reduced occurrence of crime
(Donovan & Prestemon, 2012; Hayden-Gilstad et al., 2015; Kuo et al., 2008; Kuo et al.,
1998; Troy et al., 2012; Wolfe et al., 2012). Although cities may recognize the
importance of green spaces on human health (Levitz et al., 2014), little is known about
the impacts on human behavior specifically in reference to the motivation to commit or

4

consider committing crimes. Therefore, further research is necessary to examine the
potential of this relationship.
An understanding of this research and theoretical basis behind the relationship
between mental health, criminally-motivated behavior, and the potential of urban trees to
reduce crime could support cities in urban greening efforts, regulatory policies or
protocols, and support healthier communities.

5

Chapter 2: Literature Review
2.1 Introduction
The following review explores the theoretical basis of: 1) nature’s influence on
psychological and physical health, and 2) criminally motivated behavior. Synthesis of
these theories provides a framework for understanding how urban forests have the
potential to reduce crime occurrence by alleviating the precursory mental conditions that
lead to unlawful behavior. Ultimately, this research proposes that incorporating urban
forestry practices into urban infrastructure and design will provide strategies for
evidence-based city planning and crime reduction efforts.

2.2 Physiological and Psychological Benefits of Exposure to the Natural
Environment
2.2.1

Discussion of Theory
The technological advancements of modern day life are thought to promote a

new, widespread societal illness: stress (Bratman, Daily, Levy, & Gross, 2015; Grahn &
Stigsdotter, 2003). The outcomes of stress, which may act either as a symptom or a cause,
present themselves both physically and psychologically. In the well-known fight or flight
theory, the physical process involves the brain receiving messages about the surrounding
environment, assessing situations for threat or urgent action, and priming the body to
react through a series of hormonal signals. Grahn & Stigsdotter (2003) use the biological
basis of the fight or flight theory to articulate the psychological impacts on the body,
describing present-day stress as a condition of disharmony between “what we are able to

6

accomplish and what is demanded of or expected from us, [leading] to a feeling of being
unable to control our [lives]” (p. 3).
The connection between mind and body explains why the myriad causes
demanding attention typically result in mentally fatigued and physically depleted
individuals. Mental fatigue affects executive functioning skills, namely: 1) working
memory – the ability to mentally hold and manipulate information such as when solving
problems or setting goals; 2) cognitive flexibility – the ability to shift rules and switch
tasks; and 3) inhibitory control – the ability to resist temptation or delay gratification (A.
Diamond, personal communication, February 21, 2013). The negative impacts to
executive functioning, also referred to as cognitive processing, present as “thoughtless,
tactless, and unstrategic” actions (Kuo & Sullivan, 2001b, p. 546), as well as lack of
inhibitory control and impulsive behavior (Kaplan, 1995). Emotionally, individuals
experience irritability and then frustration, potentially leading to aggressive behavior if
left unchecked (Grahn & Stigsdotter, 2003; Kuo & Sullivan, 2001b).
Stephen Kaplan’s Attention Restoration Theory (ART) (1995) provides a wellarticulated model of the prevailing theories regarding stress-related physiological impacts
and nature. Building upon James’ (1892) initial division between voluntary attention (as
cited in Berman, Jonides, & Kaplan, 2008, p. 1207), where cognitive processes control
attention, and involuntary attention, where engaged attention comes without directed
effort, Kaplan introduced the theory that nature allows for restoration because it does not
demand our voluntary, or directed, attention. Directed attention requires application of
executive functioning skills and, as noted above, without periods of repose individuals
will eventually suffer from exhaustion and loss of inhibitory control. Kaplan (1995)

7

refers to the involuntary, non-cognitive capacity as “fascination.” His proposal suggests
that fascination takes over and provides space for mental restoration in nature settings
because this environment mitigates or soothes the demands on our cognitive faculties
caused by modern society.
Paralleling the idea of nature’s power of restoring attentive and cognitive
capacity, Ulrich’s (1984) Stress Recovery Theory (SRT) proposes that nature is
conducive to mental health because it reduces stress and therefore allows for recovery.
The logical basis draws upon the theory of fight or flight stress responses, positing that
the brain must quickly assess situations for potential threats. Nature sends messages of
safety or absence of threat, creating periodic spaces for restoration and recovery.

2.2.2

Theoretical Synthesis and Application
Synthesizing the SRT and ART models, Hartig et al. (1991) demonstrated that

while the ART model focuses on the cognitive response and the SRT model on the
emotional response, both contribute to an understanding of how nature supports relief
from mental fatigue, facilitating mental restoration. Additionally, both theories agree that
a view of nature provides more benefit than other senses, and direct experience provides
the next highest benefit. The literature largely supports that the ability to see nature
represents the greatest influence over lowering stress levels (Ulrich, 1984; Leather,
Pyrgas, Beale, & Lawrence, 1998; Maas, Verheij, Groenewegen, de Vries, &
Spreeuwenberg, 2006), and may subsequently encourage outdoor experience and more
visits to natural areas (Grahn & Stigsdotter, 2003; Velarde et al., 2007).

8

The closeness of nearby nature, such as an urban park or trail, facilitates ease of
access and more frequent use; the latter is positively associated with lower stress levels
(Grahn & Stigsdotter, 2003) and improved perceptions of overall health (Maas et al.,
2006). The argument follows that in urban areas where technology overruns our
attentional capacity, including green space (parks, trees, or other vegetation) provides
benefit to human health by relieving stress and mental fatigue, and an escape from our
modern, day-to-day city life. The well-documented physiological and psychological
benefits of including nature via green spaces in cities lead to increases in positive
attitudes and higher community satisfaction (Sanesi et al., 2006), improvement in
cognitive functioning and positive emotional states (Berman et al., 2008; Bratman et al.,
2015; Hartig et al., 1991; Kaplan, 1995), and decreases in stress levels and faster stress
recovery times (Grahn & Stigsdotter, 2003; Kaplan, 1995; Ulrich, 1984). Additionally,
more green space (by volume and proximity) clearly demonstrates a positive, stressreducing impact (Maas et al., 2006; Wolfe & Robbins, 2015).

2.3 Criminal Behavior and Motivations in Urban Areas
2.3.1

Discussion of Theory
While researchers and city officials may recognize the importance of green spaces

on human health (Levitz et al., 2014), questions remain about the impacts on human
behavior, specifically in relation to the motivation to commit or consider committing
crime. The literature relating criminal behavior to urban green space references two
broadly implemented models – the “Broken Window Theory” and the “Routine Activity
Theory” (Donovan & Prestemon, 2012; Cohen & Felson, 1979; Becker, 1968).

9

The Broken Window Theory proposes that unmaintained areas represent zones of
public disorder and are therefore conducive to continued and increased criminal activity
(Wilson & Kelling, 1982). Following this model, certain environmental conditions—the
presence of litter, gangs of youth, overgrown or unruly vegetation, vandalism, graffiti,
publicly intoxicated individuals, homeless individuals, and open drug use--promote a
perceived lack or absence of social control. These “smaller” crimes, commonly termed
“social incivilities,” set the stage in these areas for larger, more aggressive crimes to
occur, such as theft, assault, rape, and murder (Wilson & Kelling, 1982).
Despite receiving some criticism and needing further research-based validation
(Center for Evidence-Based Crime Policy, 2015), the Broken Window model continues
to serve as a focal point (Sousa & Kelling, 2006) as demonstrated by the past three
decades of policing practices and crime prevention strategies. Indeed, Sousa and Kelling
(2006) found that law enforcement actions may maintain the public order by increasing
arrests for smaller, misdemeanor crimes leading to an associated decrease in larger,
felony crimes (as cited in Wilson & Kelling, 2006, p, 171).
The Routine Activity Theory outlines three preliminary criteria that need to be in
place for a crime to occur: a potential criminal, victim, and regulatory authority (Cohen &
Felson, 1979). Essentially, there must be an opportunity to commit a crime, a benefit to
the criminal of some kind, and a sufficient lack of authoritative oversight for a criminal to
consider the act. If the surrounding environment is devoid of regulatory enforcement, a
criminal with opportunity will have greater incentive to commit an offense. Additionally,
a criminal will assess the potential reward, whether pecuniary or not, as well as the
consequences before committing a crime (Donovan & Prestemon, 2012).

10

2.3.2

Factors in Urban Crime
While crime in urban areas results from multiple factors, the Broken Window and

Routine Activity theories provide a basis for understanding the geographical occurrence
of crime and the general motivations for committing crime: 1) opportunity, 2) intent, 3)
rewards, and 4) consequences. Similar to the Broken Window Theory, territoriality
(defining ownership and respect of space for protection purposes) and the defense of
property (fences, bolts, locks, etc.) play a role in a criminal’s target selection process and
are linked to reduced crime (Brown & Altman, 1983) by giving cues to the level of owner
and neighbor concern and/or vulnerability in an area (Brown & Bentley, 1993).
Studies that assess the relationship between environmental design and nature in
urban areas and criminal activity also use census block group data to address population
density, racial/ethnic composition, median income, education levels, and vacancies in
rental housing (Donovan & Prestemon, 2012; Gilstad-Hayden, 2015; Kuo & Sullivan,
2001a; Troy et al., 2012; Wolfe & Mennis, 2012). In a study of New Haven, Connecticut,
Gilstad-Hayden et al. (2015) found that neighborhoods with lower income and education
levels did not necessarily equate to higher crime rates, even though increased crime rates
are common in disadvantaged neighborhoods.
Other potential factors contributing to the occurrence of crime in metropolitan
areas include city size, pecuniary returns, and low arrest rates – all positively associated
to higher instances of crime over rural areas (Glaeser & Sacerdote, 1999). Whether lower
crime rates in rural areas results from an abundance of and/or easier access to nature is
currently an understudied area of research.

11

2.4 Synthesis: Known Studies Linking Criminal Behavior and Mental
States Affected by Nature
2.4.1

Urban Green Space and Crime: The Conflict
The available literature regarding urban nature and crime often describes a

dichotomous relationship, where evidence either points to a positive (inducing) or an
inverse (reducing) association between the effects of vegetation on crime occurrence
(Kuo & Sullivan, 2001a; Troy et al., 2012; Wolfe et al., 2012). The two main views,
described in the following section, both specifically assess vegetation or trees in
reference to the occurrence of crime; however, the role of the aforementioned green
spaces led to debate as to whether that space promoted or reduced criminal activity.
The prevalent positive association (crime-inducing) theory rests on research
demonstrating that low-level, dense vegetation provides cover for criminal activity,
obstructs or obscures views of potential victims and law enforcement agents, and leads to
fear of crime and low perceptions of safety (Donovan & Prestemon 2012; Hur & Nasar,
2014; Jansson et al., 2013). Nasar and Fisher demonstrated that vegetation should be
removed or methodically scaled back to reduce potential cover for criminal activity (as
cited in Kuo & Sullivan, 2001a, p. 344). Landscaping and maintenance regimes also
significantly influence perceptions of safety; this plays a role in the public use of parks,
open space, and other urban natural areas where disorderly or dense vegetation leads to
fear of crime (Ulrich, 1986; Kuo, Bacaicoa, & Sullivan, 1998; Kuo & Sullivan, 2001a;
Nasar, Fisher, & Grannis, 1993). The spacing of trees and surrounding vegetation next to
a pathway may obstruct a user’s view or line of sight, promoting fear of crime and
inhibiting use (Jansson et al., 2013). It may also be that other factors affect public use in
closed pathways, such as lighting or time of day.
12

The opposing research that describes an inverse (crime-reducing) association
between vegetation and crime occurrence demonstrates that landscapes designed with
open spaces and unobstructed views promote perceptions of safety (Kuo & Sullivan,
2001a; Sreetheran & Van den Bosch, 2014). In fact, greater amounts of vegetation
support decreases in total violent and property crimes (Kuo & Sullivan, 2001a); the
caveat is that the study focused specifically on the influence of grass and high-canopy
trees. The benefit of including green spaces in such a fashion was associated with
reduced crime, largely in part because it did not include the dense, tangled, fear-inducing
type of vegetation factored into other studies. In an urban setting, Garvin, Cannuscio, and
Branas (2013) experimented with vacant lots prone to crime; by greening these areas and
transforming them into well-maintained areas by adding vegetation or park-like elements,
they found an association with reduced crime occurrences. Although Garvin et al. did not
specifically reference or indicate a relationship, their results align well with the Broken
Window Theory – a vacant, disorderly lot made orderly by increasing vegetation
promoted public, reduced criminal activity and decreased prevalence of social incivility.
It seems further application of Broken Window Theory in research using green spaces as
a variable would support the proposal that order reduces crime.

2.4.2

Urban Green Space, Mental Fatigue and Crime: Synthesis and the
Importance of Trees
Emergent research may not always reference Broken Window Theory

specifically; however, using green space as part of the equation in crime reduction
solutions continues to gain ground. The feasibility of such a solution builds upon the
evidence-based research, where: 1) severe mental fatigue may lead to violent or
13

aggressive behavior, 2) a view of or contact with nature relieves stress and reduces
mental fatigue, and 3) altering vegetation maintenance regimes to improve safety
perceptions and encourage public use may support crime suppression.
Based on a review of the available literature, the research demonstrating reduced
crime due to vegetation as a primary factor was initially documented in a series of studies
of urban public housing developments in Chicago, Illinois. Kuo & Sullivan (2001b)
broadened their exploration of the contributing factors of crime noting that the physical
environment is directly associated with violent behavior. By examining the actual
occurrence of crime documented in police reports (not just the perception or fear of
crime) and the relationship between various levels of vegetation surrounding residential
area, the duo produced rather ground-breaking results (2001a). The greenness of the
buildings’ surroundings showed an inverse relationship to crime occurrence. Growing
evidence in the ensuing years continues to show that vegetation can have a positive
impact on crime reduction (Wolfe et al., 2012).
The key component, although not identified as such in the research of Kuo and
Sullivan, was that trees and grassy areas were the vegetation types used in the analysis.
Later studies expanded upon this concept, additionally drawing from nature and mental
health research, to demonstrate that trees have the potential to reduce levels of negative
stress and may ultimately suppress criminal behavior (Donovan & Prestemon, 2012;
Hansmann, Hug, & Seeland, 2007; Grahn & Stigsdotter, 2003; Kaplan, 1995; Park et al.,
2008; Ulrich, 1976). Consistent with the findings on the type/form of vegetation assessed
in previous studies, trees must be well-maintained to provide the intended crime-reducing
benefit, and perhaps even of a certain height, crown size, and distance from other trees.

14

Donovan and Prestemon (2012) studied crime in an urban residential
neighborhood in Portland, Oregon, refining the definitive benefit of trees, revisiting the
dichotomous argument between vegetation as a fear-inhibitor versus a fear-inducer.
Smaller trees block view distance and, when planted closely to residences, actually seem
to promote crime occurrence. Alternatively, trees of a certain crown height and distance
from buildings (thus non-view-obstructing) support crime suppression, especially when
planted in the public right-of-way (Donovan & Prestemon, 2012; Troy et al., 2012).
Gilstad-Hayden et al. (2015) conducted the most recent study of trees’ impact on crime
reduction. With the main focus of validating the results of previous, similar studies, the
study’s consistent findings demonstrate that greater canopy cover has an inverse
association with crime, leading to lower crime rates. As cited earlier, greater amounts of
green space in urban areas has a positive association with reduced mental fatigue,
supporting the parallel theory that greater canopy supports crime suppression.
But why would trees reduce crime, either as a sole factor or a primary factor over
other vegetation types? Reasons may include increased safety perceptions from greening
urban areas (Garvin et al., 2013), as greater public use provides a form of regulation
beyond policing efforts. The informal surveillance resulting from broad community use
supports this theory (Coley, Kuo, & Sullivan, 1997; Jacobs, 2008; Troy et al., 2012; Kuo,
Bacaicoa, & Sullivan, 1998). Urban green spaces with trees invite broader community
use, and the additional mechanism of social regulation deters criminal behavior and
contributes to community building by strengthening social relationships and perceptions
of safety in public areas.

15

2.5 Conclusion: Gaps in Research and Advancement of the Field
Of the research available, five studies over fifteen years specifically addressed
vegetation as it relates to crime; four of these addressed trees specifically (Donovan &
Prestemon, 2012; Gilstad-Hayden et al., 2015; Kuo & Sullivan, 2001a; Troy et al., 2012;
Wolfe et al., 2012). Each study consistently noted important confounding factors,
including them as control variables when available and as appropriate. Census block
group data facilitated control for median income, race/ethnicity, education level, and
population density. The most recent study by Gilstad-Hayden et al. (2015) advocated for
the use of FBI UCR codes as a basis for further research in order to promote consistency
and comparison of results.
In an area of relatively new research, there remains opportunity for further
development of concepts and refinement of theory. This thesis research intends to address
two noted gaps by examining: 1) a core urban area where crime is concentrated versus a
city or county-wide assessment, considering the higher rate of crime in urban versus rural
areas; and 2) a longitudinal analysis of crime occurrence over time as it relates to trees
(per my own review of existing research, no such analysis has been completed).

16

Chapter 3: Methods
3.1 Aim and Objectives of Research
The following research focuses on validating the relationship between urban tree
canopy cover and crime occurrence, specifically exploring the hypothesis that a greater
percentage of tree canopy cover in an urban core area has an inverse effect on crime,
leading to reduced instances. To accomplish this main objective, a defined urban space
was established as the study area; socioeconomic variables at the census block group
level were controlled for and assessed; statistical and Geographic Information System
(GIS) analysis was used to examine the potential effect of urban tree canopy cover on
crime occurrence; and the inclusion of spatial data and analysis helped to mitigate the
impact of surrounding areas on the imposed study area boundary.
Examining a potential temporal relationship between canopy cover and crime
instances using a repeated cross-sectional analysis to assess potential change over a
seven-year time span will be used to explore the relationship in more depth. This
secondary hypothesis proposes that as more trees are planted or as trees grow larger over
time, providing a greater urban tree canopy cover, reduced crime occurrence will follow.
Testing both hypotheses involved data collected from third-party sources and data
generated through the use of ArcGIS software.

3.2 Study Site Description
The City of Olympia is located in Thurston County, Washington at the southernmost tip of Puget Sound on Budd Inlet. With a population of 49,670 as of 2014, Olympia
encompasses approximately 19.72 square miles (City of Olympia, 2015). The police
17

department divides the City of Olympia into six patrol sectors, where sector “B”
encompasses the downtown business core, extends up the west and east sides of Budd
Inlet, up to and including the State Capitol Campus (see Figure 1). The City of Olympia
defines a comparable area, with respect to boundaries, as the “Downtown Strategy Area,”
although it excludes the Capitol Campus area (see Appendix A).
To meet the intent of this analysis, patrol sector B provides full coverage of the
urban area of interest and several bordering, semi-residential areas. Additionally, using
the patrol sector as the selected study area serves three purposes:
1) Previous studies focused on broad or varied geographical areas, at single
neighborhood, city-wide, and county-wide scales. In order to examine the
impacts of urban forest canopy cover specifically on crime occurrence, this
study focuses on an urban core and, to account for border effects, the areas
closely surrounding that core.
2) Olympia dedicates a significant amount of resources and coordination to
providing a comprehensive plan for improving the downtown area in terms of
increased safety, greater economic revenue, broader community participation,
and support for natural and historic spaces (City of Olympia, 2015). Focusing
on patrol sector B supports Olympia’s downtown strategic actions and will
provide data to inform funding decisions, tree maintenance regimes, and
safety regulation protocols.
3) As cities begin to address ways to account for the impacts of climate change,
increased urban populations, and the need for accessible open green space,
support for urban “greening” methods continues to increase. Further data to

18

support the inclusion or development of additional green space within cities,
specifically trees, can contribute to decisions made that affect overall public
health and environmental well-being.

Figure 1. The defined study area based on the City of Olympia Police Department
boundaries for patrol sector “B,” depicted over a map of Olympia, Washington.

19

3.3 Data
3.3.1

Variables

3.3.1.1

Urban Tree Canopy Cover
Canopy cover measures the tree crown diameter combined across all trees in a

designated area; some studies cite tree crown radius as an alternate measurement
(McPherson et al., 2005). Canopy cover need not be continuous by this definition; rather,
the total percentage of coverage accounts for the measurement in the designated area.
Percentage of tree canopy cover provides the necessary variable for this analysis, based
on research demonstrating that a view of trees over other vegetation, high-canopied trees,
and clear, open areas with well-spaced trees provides maximum, positive psychological
benefit.
For this study, an urban tree canopy cover dataset was created using Light
Detection and Ranging (LiDAR) and orthoimagery (high resolution aerial images with a
near-infrared band) data obtained from the City of Olympia. LiDAR is a remote-sensing
technique that uses laser light pulses to survey and model the earth’s surface, often
through aerial methods. The laser pulses reflect on the earth’s surface and objects on the
earth’s surface (buildings, trees, etc.), returning points that can be geographically
referenced for analysis.
During 2008 and 2015, the City conducted two LiDAR data collection events.
Data for 2015 were collected to “match” the specifications of the 2008 data set,
producing comparable results. The City provided raster data for each year in the form of
an ArcGIS file geodatabase, using 3.0 Foot Esri Grids (LiDAR Bare Earth and Highest
Hit models that reference the lowest and highest laser return points, respectively) and 1.5

20

Foot GeoTiffs (high-intensity orthoimages). See Appendix B for visuals of the raw
datasets.
3.3.1.2

Crime Incident Data
Crime rate figures in both rural and urban areas originate from data tracked

voluntarily by many cities, counties, and states. The Uniform Crime Reporting (UCR)
Program, managed by the Federal Bureau of Investigation (FBI) since 1930, collects this
nationwide crime data for statistical purposes. The most recently published data for the
United States demonstrates a rate of 395.7 violent crimes per 100,000 inhabitants in
urban areas, as compared to national violent crime rate of 375.7 (Uniform Crime Reports,
2015). The violent crime rates in Washington State and the City of Olympia are 285.2
and 198, respectively (Uniform Crime Reports, 2015). Considering that the this thesis
research examines a core urban area, it is additionally useful to note that the national total
crime rate outside of metropolitan areas is lower, averaging 168.5 for counties with
populations over 25,000, such as Thurston County. Drawing an exact comparison to the
City of Olympia presents challenges, as total crime rates are reported per 1000 people,
versus 100,000. The Olympia Police Department reported total rime rates of 53.5 in 2008
and 86.3 for 2014 (see 3.4.3 Limitations, Crime Coding Effects for further information
about the large gap in reported crime rates).
For this study, the City of Olympia Police Department provided annual crime data
for 2008 and 2014 in the format of two Excel spreadsheets1. The data included date and
time of the crime incident, type of crime, and geo-coordinates based on the North
American Datum (NAD) 1983 State Plane for Washington (South FIPS, 4602, US Feet).

1

The City of Olympia provided the annual crime data in response to a public records request submitted by
the author.

21

This commonly used projected coordinate system supports best-practice with highlylocalized, georeferenced data and provides detailed accuracy (with errors less than
1:10,000).
3.3.1.3

Socio-demographic Data
Following the methodology established for control variables in correlation studies

between trees and crime (Gilstad-Hayden et al., 2015; Troy et al., 2012; Wolfe &
Mennis, 2012), socio-demographic data was sourced from the U.S. Census Bureau’s
TIGER (Topologically Integrated Geographic Encoding and Referencing) database, using
shapefiles prejoined with Amercan Community Survey (ACS) demographic data tables at
the census block group level. The selected ACS shapefiles included the five-year detailed
estimates for the 2008-2012 and 2010-2014 surveys to most closely align with the
available LiDAR/orthoimagery and corresponding crime data in this study.
The socio-demographic variables utilized for this study included a subset of the
controls selected in previous studies (noted in preceding paragraph), namely median
income (median household income in the past twelve months based on 2012 and 2014
inflation-adjusted dollars, respectively), education level (percentage of the area
population without at least a high school diploma), and race/ethnicity (percentages of the
area population identifying as Hispanic/Latino and as Non-Hispanic/Latino and African
American/Black), and total population per block group within the study area.
3.3.1.4

Spatial Data
Census block group boundaries provided the necessary geographical parameters

for spatial referencing and analysis of the established dependent and independent
variables. Census tracts are divided statistically into block groups, which are designated

22

to contain approximately 600 to 3000 people (United States Census Bureau, 2012). Block
group data were sourced from the U.S. Census Bureau’s TIGER database, using the same
selection as described for the socio-demographic data. The data has been transformed by
Esri into ArcGIS compatible shapefiles, available for download and direct import into
ArcMap 10.3 (see Figure 2).

Figure 2. Census block group boundaries overlaying the downtown core of Olympia,
Washington. Block group boundaries from the 2008-2012 dataset are depicted, although
the 2010-2014 dataset had identical boundaries.
23

3.3.2

Data Preparation
Preparation of each data component – urban tree canopy cover, crime, and

socioeconomic factors – involved some initial refinement using Microsoft Excel 2010
and JMP Pro 12.1.0. However, the majority of the data preparation involved curation via
Geographic Information Systems (GIS) by either creating the datasets within, or
importing them into, the ArcGIS software suite and compiling a spatial database. Using
ArcCatalog and ArcMap 10.3 for refinement of each variable contributed to alignment of
all data at the census block group level within the study area.
To derive percentages of urban tree canopy cover, the LiDAR and GeoTiff
orthoimagery 2008 and 2015 datasets were imported into ArcMap 10.3 and clipped to
cover the designated study area. Exploratory analysis was performed using the bare earth
(BE) and highest hit (HH) LiDAR data to create a shaded relief (also called hillshade)
map of the study area, assessing for general elevation differences and potential basins of
vegetation that would not be defined as canopy cover. This method assessed vegetation
levels above five feet in height and within the highest return point range: HH - (BE + 5
feet). Next, two urban tree canopy layers were digitized and edited using the
orthoimagery for each year. The editing process involved drawing polygons over the
orthoimagery, each of which constituted a unique feature in the GIS layers’ attribute
tables (see Figures 3 and 4). Data from each feature was summed and organized by
census block group boundaries overlaying the study area. The boundaries provided the
parameters for determining the percentages of urban tree canopy coverage, calculated
from the attribute table for each year’s layer.

24

Figure 3. Tree canopy layer created using ArcMap 10.3 from 2008 orthoimagery.

25

Figure 4. Tree canopy layer created using ArcMap 10.3 from 2015 orthoimagery.

26

From the Excel spreadsheets for each year, crime data was coded according to the
National Incident-Based Reporting System (NIBRS), managed by the Federal Bureau of
Investigation (FBI) Uniform Crime Reporting (UCR) Program. However, due to
discrepancies between crime data reporting methods for 2008 and 2014 (discussed in
3.4.3 Limitations, Crime Coding Effects), the variables calculated for analysis included
total crime counts (per incident), crime density, and total crime rates versus
categorization by crime type. Each crime dataset was geocoded and imported into
ArcMap 10.3 and variables were calculated within the census block group boundaries
that overlaid the study area.
Socioeconomic data, coming in geographically-referenced tables for all of
Washington State, were first narrowed to the relevant census block groups covering the
study area using ArcMap 10.3. Using Excel, percentages were calculated for the selected
control variables: median income, race/ethnicity, educational attainment, and population
density. Since the census block group is a relatively small geographic area, the
calculation for population density used square kilometers instead of square miles. All
control variable calculations were spatially joined to the corresponding block group.

3.4 Limitations
3.4.1

LiDAR data
While the LiDAR data collection events for 2008 and 2015 covered the same

geographic area (the City of Olympia), the exact time of year that the aerial flight took
place does not appear to perfectly align. The 2015 aerial flight took place in late May.
The month during which the 2008 flight took place is unknown; however, based on

27

observational assessments of Capital Lake in the GeoTiff files, and the presence of algal
blooms which typically occur from May through October, it appears the flight took place
later in the summer. Similar algal blooms do not appear to be present in the 2015
orthoimagery dataset. The trees in both datasets appear to be in full leaf-out stage,
lending support to the conclusion that the time discrepancy likely will not significantly
affect the calculations of percent tree canopy coverage for respective years.
Additionally, there were some challenges in discerning the boundary of tree
crowns when densely surrounded by vegetation. Assessing shaded relief maps using the
BE and HH datasets provided some mitigation of error, and the infra-red band of the
orthoimagery supported further clarity in discerning shadows, vegetation, and taller
shrubbery from tree canopy. Based on these methods, the error in identifying tree canopy
is estimated to be minimal.
Finally, analysis for this research began in mid-2015, as the most recent LiDAR
event occurred. To secure corresponding crime and socio-demographic data at an annual
scale, 2014 was selected as the most recent, full year for data analysis. Since primary tree
growth occurs during the winter, the time delay will likely not have a large impact in this
analysis.

3.4.2

Variable Boundaries
The main study area boundary was provided by the Olympia Police Department’s

patrol sector “B”; however, the two of the census block groups’ boundaries partially
extend into patrol sector “A.” Since crime data was not sourced for sector “A,” the crime
density variable for this block group is expected to be lower when using only sector “B”

28

crime data. Considering the relatively low incident rate (see Appendix C, map legend
inset) and that canopy coverage is identified only within patrol sector “B,” it may actually
be improve the accuracy of the analysis to isolate the B sector crimes; including
additional crime data would likely inflate the results and lead to incorrect inferences.

3.4.3

Crime Coding Effects
Between the 2008 and 2014, the years of selected crime data for this study, the

State of Washington (and therefore the City of Olympia) changed the way such data are
collected and reported. In 2008, the State used the Washington Uniform Crime Reporting
(UCR) Program, based on guidelines and a national UCR program established by the
Federal Bureau of Investigation (FBI). Due to program inefficiencies and lack of
sufficient detail in the Summary UCR data collected, the State opted to transition to the
National Incident-Based Reporting System (NIBRS) which collects more in-depth data
on individual crime occurrences (Washington Association of Sheriffs and Police Chiefs,
2008). The 2014 data represents the third year in which the State captured crime
occurrence using the NIBRS coding guidelines.
For this study, the impact of the change from Summary UCR to NIBRS methods
means that the crime count for 2014 may be slightly inflated, giving the appearance of
more crimes having occurred, or may appear to have a more significant increase over
2008 occurrences than is warranted. Multiple “incidents” may be reported for each crime
and coded separately in NIBRS, reflecting the difference in how crimes are categorized.
To mitigate this impact, the study’s analysis used total crime, crime rate, and crime

29

density calculations versus previously established variables for distinguishing violent or
property crimes, as categorized under the Summary UCR guidelines.

3.5 Statistical Analysis
Analysis began by assessing the characteristics of the crime, urban tree canopy
cover, and socio-demographic data, using descriptive statistics of the mean, standard
deviation, and range for each variable. In ArcGIS 10.3, an exploratory analysis was
conducted to examine the spatial distributions of crime and identify potential patterns or
clustering. This method of data visualization used crime heat maps and an underlying
choropleth layer shaded by tree canopy distribution.
The first hypothesis, examining urban tree canopy cover and crime occurrence,
was tested using Ordinary Least Squares regression, which can be expressed as:

Y = β0 + β1X1 + β2X2 + … βnXn + ε
Initial regression models included the variables total crime count, crime rate, and crime
density and urban tree canopy cover, but not the control variables. Due to theoretical
considerations (based on subsequent spatial regression analysis described later) and
regression assumptions not being met using crime count or density calculations, a fitted
regression model using crime rate as the sole outcome variable was selected for each
year. For this model, histograms and residual error plots demonstrated more normally
distributed error terms and reductions in heteroscedasticity, respectively. To incorporate
the four control variables, the sample data aggregated at the block group level (n=6) was
bootstrapped (n=6000) to meet regression assumptions. Variance inflation factors were

30

persistently high when incorporating race and ethnicity (VIF > 4), indicating
multicollinearity. Ethnicity proved a better fit than race, based on model fit assumptions
and a lower VIF factor. Similar results occurred with median income and educational
attainment (VIF > 41), however, due to model over-fitting, these variables were also
removed.
Following OLS regression for each year, Global Moran’s I tests were carried out
on the residuals to assess for spatial dependency; each resultant statistic demonstrated
significant spatial autocorrelation. This effect results from block groups sharing physical
boundaries with other block groups and city zones, affecting the assumed independent
nature of the data. As a result, the independent variables and error terms cannot be
assumed to be uncorrelated and the inferences may be overestimated if not accounting
for spatial dependency, i.e., crime is affected by variables in neighboring block groups.
Crime occurrence tends to be relatively the same in neighboring block groups versus
those further away due to spillover effects (Gilstad-Hayden et al., 2015; Troy et al., 2012;
Wolfe & Mennis, 2012).
To address the effects of spatially dependent data, spatial analysis takes into
account arbitrary boundaries and effects of near-boundary incidents. A spatial weights
matrix was constructed, specifically a k-nearest neighbors matrix which places weights
on each variable according to the distance between a value and the number (k) of nearest
neighbor values. The block group shapefiles joined with explanatory variables were
imported into ArcGIS 10.3 to create the matrix and. Including the spatial weights matrix
as part of the Global Moran’s I test allows for measurement of the linear relationship

31

between a variable and the weighted average of its neighbors, using a model that can be
expressed as:

𝑛 ∑𝑛𝑖=1 ∑𝑛𝑗=1 𝑤𝑖,𝑗 𝑧𝑖 𝑧𝑗
𝐼=
𝑆0
∑𝑛𝑖=1 𝑧𝑖2
where i and j represent features (values of observations), n is the total number of features,
zi and zj denote the deviation from the mean of a feature, wi,j denotes the weighted value
between two features (i and j) of the row-standardized matrix, and S0 is the sum of all
weights (wi,j) in the matrix. Using this model, with an incorporated spatial weights matrix,
Global Moran’s I tests were re-run on the OLS residuals.
The second hypothesis, assessing the change in tree canopy cover compared to
crime occurrence over time, was tested using a repeated cross-sectional analysis. The
same percentages of block group canopy cover and crime rate calculations were
organized by year and imported into ArcGIS 10.3 to assess mean change over time and
examine spatial differences.

32

Chapter 4: Results
4.1 Descriptive Statistics
Table 1 provides the descriptive characteristics of each regression variable,
summarized by block group and year of analysis. The dependent variable (crime rate) and
independent variables (tree canopy, socio-demographic factors) demonstrated clear
variability across block groups. Total crime and total population are included as reference
for the calculations of other variables. Averages for Total crimes and Crime rate were
225 and 385.35 in 2008 and 178.33 and 249.61 in 2014, respectively2. Tree canopy
coverage for both years remained fairly steady at approximately 13%, with a slight (less
than 1%) increase in 2015. The average median income for household block groups
decreased in 2014, although it was interesting to note the shift in range maximum by
approximately $12,000. Total populations within block groups varied slightly each year,
but percentages of non-Hispanic African-American/Black and Hispanic-Latino remained
relatively the same.
Examination of heat and choropleth maps revealed strong clustering of crime in
less densely canopied areas for both years, suggesting that block groups with less urban
tree canopy cover experience less total crime occurrence (see Figure 5).

2

Crime rates appear inflated over those reported by the Olympia Police Department due to calculations
conducted on population levels under 1000. For consistency in results, the numbers were calculated similar
to standard reporting protocols.

33

Table 1. Descriptive statistics for study area block groups in Olympia, Washington.
Variable

Mean

Std. Dev.

Minimum

Maximum

2008 (n = 6)
Total crimes

225.00

424.33

0.00

1071.00

Crime rate (per 1000 people)

385.35

806.43

0.00

2020.75

13.07

14.24

0.44

25.45

58648.17

35345.84

14884.00

98228.00

Educational attainment (% with
less than high school diploma)

1.97

2.59

0.00

6.79

Hispanic-Latino (%)

5.25

5.71

0.00

16.01

Non-Hispanic, AfricanAmerican/Black (%)

1.83

3.54

0.00

8.85

Total population

924.17

370.25

530.00

1605.00

Population Density
(1000 people/sq km)

429.70

239.90

95.67

724.26

Total crimes

178.33

313.60

1.00

802.00

Crime rate (per 1000 people)

249.61

473.85

0.66

1206.02

13.50

14.23

0.58

33.50

52964.50

27444.10

16591.00

86161.00

Educational attainment (% with
less than high school diploma)

1.40

1.37

0.00

3.46

Hispanic-Latino (%)

5.65

5.06

0.00

14.01

Non-Hispanic, AfricanAmerican/Black (%)

1.85

2.31

0.00

5.38

Total population

912.00

328.13

640.00

1523.00

Population Density
(1000 people/sq km)

420.91

233.03

120.04

724.26

Tree canopy cover (%)
Median Income

2014 (n = 6)

Tree canopy cover (%)
Median Income

34

Figure 5. Choropleth map of 2008 block groups shaded by percent canopy cover. 2015
choropleth maps demonstrated similar results.

4.2 Statistical Analysis Results
Table 2 provides the results from the Ordinary Least Squares regression analysis
assessing the explanatory power of urban tree canopy cover and socio-demographic
35

variables on total crime rate for both 2008 and 2014-15 datasets. Urban tree canopy cover
had a statistically significant relationship with crime rate in 2008 (β = -13.114, p < 0.001)
and in 2014 (β = -7.910, p < 0.001).

Table 2. Results from Ordinary Least Squares regression models.
Crime Rate (per 1000 people)
β

SE

Pr > |t|

-13.114

0.591

< 0.001

Median Income

n.s.

n.s.

n.s.

Educational attainment
(% with less than high school diploma)

n.s.

n.s.

n.s.

-63.009

1.282

< 0.001

n.s.

n.s.

n.s.

-2.001

0.033

< 0.001

1746.414

16.417

< 0.001

-7.910

0.302

< 0.001

Median Income

n.s.

n.s.

n.s.

Educational attainment
(% with less than high school diploma)

n.s.

n.s.

n.s.

-51.242

0.811

< 0.001

n.s.

n.s.

n.s.

-0.663

0.019

< 0.001

924.644

8.639

< 0.001

Variable
2008 (n = 6000)
Tree canopy cover (%)

Hispanic-Latino (%)
Non-Hispanic, African-American/Black (%)
Population Density (1000 people/sq mile
Intercept
Adjusted R-Squared

0.57

2014 (n = 6000)
Tree canopy cover (%)

Hispanic-Latino (%)
Non-Hispanic, African-American/Black (%)
Population Density (1000 people/sq mile
Intercept
Adjusted R-Squared

0.58

36

OLS regression using ArcGIS 10.3 provides a geographically symbolized map of
the standardized residuals (see Figures 6 and 7). Following OLS regression for each year,
a Global Moran’s I test on the residuals from both 2008 and 2014 regression models
demonstrated significant spatial autocorrelation (I = 1, p < 0.0001). The Moran’s I
statistic is measured between -1 and 1, where values closer to positive 1 represent greater
spatial clustering and values are autocorrelated. After constructing a spatial weights
matrix, a second Global Moran’s I test was carried out on the residuals. The resulting
Moran’s statistic demonstrated only slight reductions in the effect of spatial
autocorrelation (I = 0.998, p < 0.0001) for both years.
While the Moran’s I test examines the potential neighboring influences of spatial
data, the Koenker (BP) Statistic describes the consistency of the relationship between the
modeled dependent variable and explanatory variables in both geographic space and data
space. With consistent relationships in the geographic space, the data are said to be
stationary. The parallel in data space is minimizing heteroscedasticity in the regression
model, where the size of the error terms varies significantly across the explanatory
variables’ values.
Further examination of the results after including the Moran’s I statistic revealed a
significant BP statistic (5772.245, df = 3, p < 0.05), demonstrating statistically significant
regional variation (non-stationarity) in the geographic properties of the explanatory
variables data. The significant BP values for both 2008 and 2014 supported using robust
standard errors over regular standard errors, as well as using the Wald Statistic
(6093.281, df = 3, p < 0.01) to assess good over-all regression model fit for analysis.

37

Cross-sectional time analysis involved spatial exploration and comparison of the
mean differences between the change in crime rate and the change in percentage canopy
cover between 2008 and 2014. While there was a relatively large decrease in the crime
rate over 7 years, the analysis demonstrated a non-significant relationship in the reduction
of crime occurrence and the approximate 0.5% increase in tree canopy cover.

Figure 6. Geographically symbolized map of residuals for 2008 analysis. Created with
ArcMap 10.3.

38

Figure 7. Geographically symbolized map of residuals for 2014-15 analysis. Created with
ArcMap 10.3.

4.3 Discussion
The City of Olympia’s urban core served as a study area to examine the
relationship between urban tree canopy cover and the occurrence of crime. The small
scale provided the opportunity to assess and validate whether greater tree canopy cover

39

supports crime suppression in primarily non-residential areas. The study area
encompassed six census block groups for which data were compiled and examined.
Regression analysis of the relationship between urban tree canopy cover and
crime occurrence, independent of socio-demographic controls, demonstrated that areas
with higher percentages of canopy cover are associated with reduced instances of crime.
The inclusion of socio-demographic controls further supported the validated inverse
relationship between canopy cover and crime occurrence by providing a mechanism to
measure potentially confounding factors in the analysis. The results support the
associated findings of other studies concerning tree canopy cover and crime suppression
at various geographic scales (Donovan & Prestemon, 2012; Gilstad-Hayden et al., 2015;
Kuo & Sullivan, 2001a; Troy et al., 2012; Wolfe et al., 2012).
While the resulting regression analysis of trees and crime agreed with the
hypothesized relationship, the limitation of certain control factors in this study warrants
further consideration. Although block groups with higher population density
demonstrated reduced crime occurrence, this may be due to the urban nature of the study
design which includes a limited population residing in the more urban census block
groups of Olympia’s defined downtown zone. This may also explain why the expected
relationships between median income and educational attainment with crime occurrence
were not significant. Without sufficient data, due to low population counts, these
variables had negligible effects and instead served to over-fit the regression model.
Upon close analysis of Olympia’s downtown, one can find a lack of open green
space and several notable areas where no trees are present for several blocks. The
argument follows that if a view of nature positively supports mental health, and positive

40

mental health leads to reduced crime occurrences, then the areas lacking trees should
have inversely higher crime rates. The results from this study align with the theoretical
basis for selecting of tree canopy cover as a variable in crime suppression studies, versus
vegetation or green spaces in general, mainly because it is not the mere presence of trees
or a specific count of trees, but the overall percentage of coverage in a given area. The
ability to view nature is enhanced specifically by trees due to the inherent nature of
height – those trees that can tower above buildings, lots, or other urban spaces are easier
to see and contribute to the greening of the visual landscape.

41

Chapter 5: Conclusion
5.1 Research Conclusions
Increasing residential populations in cities continue to accentuate the importance
of incorporating green space within urban areas. Losses in tree canopy cover contribute to
the shortfall of several beneficial factors, including the provision of mental health
benefits and suppression of crime. This study provided an opportunity to examine urban
trees at a small geographic scale, in the urban core of a moderately sized city. While
additional scaled models could provide further validation, the results from this research
demonstrate that maintaining higher percentages of tree canopy cover can support
reduced crime occurrences.
As cities balance economic and social demands, funding shifts generally have a
negative impact on park enhancement or maintenance priorities. The effects of reduced
availability or quality of green space may not understood or immediately prevalent.
However, research continues to demonstrate the importance of nature to mental health.
Providing data to support availability, access, and abundance of trees supports these
findings and may encourage cities to invest further in planting and care regimes.
“Never say there is nothing beautiful in the world anymore. There is always
something to make you wonder in the shape of a tree, the trembling of a leaf.”
Albert Schweitzer (Theologian, Surgeon, Nobel Peace Prize Winner)

5.2 Recommendations for Further Research
There are opportunities to enhance the research conducted in this study. First, a
deeper understanding of the NIBRS coding system could support categorizing crimes by
type and understanding their specific relationship to urban tree canopy cover. For
42

example, since more businesses are present in a downtown setting there may be a
relatively higher rate of property crimes such as theft or vandalism. Alternatively,
drawing a greater number of people into the space for shopping may increase the violent
crime rate due to the increased chances for robbery or assault.
A second opportunity for further analysis is the examination of longitudinal
effects. A temporal component was attempted in this study following research by GilstadHayden et al. (2015). However, due to the discrepancies in crime coding and the potential
for false inflation of the crime rates, it was not conclusive to assess the rates of change
over the seven-year span in this study. Further attempts to assess whether tree canopy
cover has a long-term relationship with crime suppression could provide insight into
urban greening methods and provide support for consistent funding and maintenance
regimes.
Additional considerations include the informal regulatory power of more citizens
utilizing open green space. Utilization of a park layer and statistics on community use for
a given area may provide insights into the relationship between open space and reduced
mental stress, as hypothesized by Kuo (2001a). The phrase “more police on the beat,
more crimes on the street” refers to the increased prevalence of arrest when more officers
are on duty. If green spaces draw in more residents, would there be a lower need for more
formal regulation and policing? And would this result in fewer arrests because officers
are not present (though not a reduction in crime), or in fewer crime incidents because of
the larger population influence? These questions and more could be analyzed using a
more in-depth study design that incorporates crime coding and a geographically weighted
regression analysis that allows for regional variation.

43

Works Cited
American Forests. “ Urban Forests.” American Forests.org. American Forests. n.d. Web.
15 May 2015.
Berman, M. G., Jonides, J., & Kaplan, S. (2008). The Cognitive Benefits of Interacting
With Nature. Psychological Science (Wiley-Blackwell), 19(12), 1207-1212.
Bratman, G. N., Daily, G. C., Levy, B. J., & Gross, J. J. (2015). The benefits of nature
experience: Improved affect and cognition. Landscape and Urban Planning, 138,
June 2015, Pages 41-50.
Brown, B. B., & Altman, I. (1983). Territoriality, Defensible Space and Residential
Burglary: An Environmental Analysis. Journal of Environmental Psychology, 3,
203-220.
Brown, B. B., & Bentley, D. L. (1993). Residential Burglars Judge Risk: The Role of
Territoriality. Journal of Environmental Psychology, 13, 51-61.
Carrus, G., Scopelliti, M., Lafortezza, R., Colangelo, G., Ferrini, F., Salbitano, F., &
Sanesi, G. (2015). Go greener, feel better? The positive effects of biodiversity on
the well-being of individuals visiting urban and peri-urban green
areas. Landscape & Urban Planning, 134(2015), 221-228.
Center for Evidence-Based Crime Policy. (2015). Broken Windows Policing. George
Mason University, 2015. Web. 13 Oct. 2015.
Choudhry, K. Z., Coles, R., Qureshi, S., Ashford, R., Khan, S., & Mir, R. R. (2015). A
review of methodologies used in studies investigating human behaviour as
determinant of outcome for exposure to ‘naturalistic and urban environments.’
Urban Forestry & Urban Greening, 14(3), 527-537.
City of Olympia. (2015). Downtown Strategy. City of Olympia, 2015. Web. 1 Jun. 2016.
City of Olympia. (2015). About Olympia – Demographics. City of Olympia, 2015. Web.
1 Jun. 2016
Coley, R. L., & Kuo, F. E. (1997). Where does community grow? The social context
created by nature in urban public housing. Environment & Behavior, 29(4), 468.
Cronon, W. “The Trouble with Wilderness.” Uncommon Ground: Rethinking the Human
Place in Nature. Ed. William Cronon. New York: W. W. Norton & Co., 1995. 6990.
Donovan, G. H., & Prestemon, J. P. (2012). The Effect of Trees on Crime in Portland,
Oregon. Environment and Behavior, 44(I), 3-30.
44

Fields, S. (2002). If a Tree Falls in the City. Environmental Health Perspectives, 110(7),
A392.
Garvin, E. C., Cannuscio, C. C., & Branas, C. C. (2013). Greening vacant lots to reduce
violent crime: a randomised controlled trial. Injury Prevention : Journal of the
International Society for Child and Adolescent Injury Prevention,19(3), 198–203.
Gilstad-Hayden, K., Wallace, L. R., Carroll-Scott, A., Meyer, S. R., Barbo, S., MurphyDunning, C., & Ickovics, J. R. (2015). Research note: Greater tree canopy cover is
associated with lower rates of both violent and property crime in New Haven, CT.
Landscape and Urban Planning, 106(2015), 248-253.
Grahn, P., & Stigsdotter, U. A. (2003). Landscape planning and stress. Urban Forestry &
Urban Greening, 2(1), 1-18.
Hansmann, R., Hug, S. M., & Seeland, K. (2007). Restoration and stress relief through
physical activities in forests and parks. Urban Forestry & Urban Greening, 6,
213-225.
Hartig, T., Mang, M., & Evans, G. W. (1991). Restorative effects of natural environment
experience. Environment and Behavior, 23, 3-26.
Hur, M., & Nasar, J. L. (2014). Physical upkeep, perceived upkeep, fear of crime and
neighborhood satisfaction. Journal of Environmental Psychology, 38(2014), 186194.
Jacobs, J. (2008). The Death and Life of Great American Cities. In B. McKibben (Ed.),
American Earth (pp. 359-364). New York, NY: The Library of America.
Jansson, M., Fors, H., Lindgren, T., & Wiström, B. (2013). Perceived personal safety in
relation to urban woodland vegetation – A review. Urban Forestry & Urban
Greening, 12, 127-133.
Kaplan, S. (1995). The restorative benefits of nature: Toward an integrative framework.
Journal of Environmental Psychology, 15, 169–182.
Kaplan, R., & Kaplan, S. (2011). Well-being, Reasonableness, and the Natural
Environment. Applied Psychology: Health & Well-Being, 3(3), 304-321.
Kirnbauer, M.C., Baetz, B.W., & Kenney, W.A. (2013). Estimating the stormwater
attenuation benefits derived from planting four monoculture species of deciduous
trees on vacant and underutilized urban land parcels. Urban Forestry & Urban
Greening, 12(3), 401-407.

45

Kuo, F. E., Bacaicoa, M., & Sullivan, W. C. (1998). Transforming inner city landscapes:
Trees, sense of safety, and preference. Environment and Behavior, 30, 28-59.
Kuo, F. E., & Sullivan, W. C. (2001a). Environment and crime in the inner city: Does
vegetation reduce crime? Environment and Behavior, 33(3), 343-367.
Kuo, F. E., & Sullivan, W. C. (2001b). Aggression and violence in the inner city: Effects
of environment via mental fatigue. Environment and Behavior, 33(4), 543-571.
Leather, P., Pyrgas, M., Beale, D., & Lawrence, C. (1998). Windows in the workplace.
Environment & Behavior, 30, 739-762.
Levitz, D. “The Role of Parks in Shaping Successful Cities: A White Paper.” National
Recreation and Park Association & American Planning Association, May 2014.
Web. 28 Nov. 2015.
Lohr, V. I., & Pearson-Simms, C.H. (2006). Responses to scenes with spreading,
rounded, and conical tree forms. Environment and Behavior, 38(5), 667-688.
Louv, R. (2012). The Nature Principle: Reconnecting with Life in a Virtual Age. New
York: Algonquin Books of Chapel Hill.
Maas, J., Verheij, R. A., Groenewegen, P. P., de Vries, S., & Spreeuwenberg, P. (2006).
Green space, urbanity, and health: how strong is the relation? Journal of
Epidemiology and Community Health, 60, 587-592.
McPherson, E.G., Simpson, J.R., Peper, P.J., Maco, S.E., & Xiao, Q. (2005). Municipal
forest benefits and costs in five U.S. cities. Journal of Forestry, 103(8), 411-416.
Nasar, J. L., Fisher, B. & Grannis, M. (1993). Proximate Physical Cues to Fear of Crime.
Landscape and Urban Planning, 26, 161-178.
Park, B. J., Tsunetsugu, Y., Ishi, H., Furuhashi, S., Hirando, H., Kagawa, T., & Miyazaki,
Y. (2008). Physiological effects of shinrin-yoku (taking in the atmosphere of the
forest) in a mixed forest in Shinano Town, Japan. Scandinavian Journal of Forest
Research, 23, 278-283.
Sousa, W., & Kelling, G. (2006). Of “broken windows,” criminology, and criminal
justice. In D. Weisburd & A. Braga (Eds.), Police innovation: Contrasting
perspectives (pp. 77 – 97). New York: Cambridge University Press.
Sreetheran, M., & Van den Bosch, C. C. K. (2014). A socio-ecological exploration of fear
of crime in urban green spaces – A systematic review. Urban Forestry & Urban
Greening, 13, 1-18.

46

Sreetheran, M., & Van den Bosch, C. C. K. (2015). Fear of crime in urban parks – What
the residents of Kuala Lumpur have to say? Urban Forestry & Urban Greening,
14, 702-713.
Sydnor, T. D., & Subburayalu, S. K. (2011). Should We Consider Expected
Environmental Benefits When Planting Larger or Smaller Tree Species?
Arboriculture & Urban Forestry, 37(4), 167-172.
Troy, A., Grove, J. M., & O’Neil-Dunne, J. (2012). The relationship between tree canopy
and crime rates across an urban-rural gradient in the greater Baltimore region.
Landscape and Urban Planning, 106(2012), 262-270.
Ulrich, R. S. (1984). View through a window may influence recovery from surgery.
Science, 224(4647), 420-421.
Ulrich, R. S. (1986). Human Responses to Vegetation and Landscapes. Landscape and
Urban Planning, 13, 29-44.
Uniform Crime Reporting Program. Crime in the United States. Federal Bureau of
Investigation, Criminal Justice Information Services, 2015. Web. 28 Nov. 2015.
U.N. Population Division. Urban/Rural Populations. United Nations, Department of
Economic and Social Affairs, 2014. Web. 24 Jan. 2015.
United States Census Bureau. (2012, December 6). Geographic Terms and Concepts Block Groups. Web. 1 Jun. 2016.
Velarde, M.D., Fry, G., & Tveit, M. (2007). Health effects of viewing landscapes –
Landscape types in environmental psychology. Urban Forestry & Urban
Greening, 6(4), 199-212.
Washington Association of Sheriffs and Police Chiefs. (2015). Crime in Washington.
Web. 1 Jun. 2016.
Washington Association of Sheriffs and Police Chiefs. (2008). Crime in Washington.
Web. 1 Jun. 2016.
Wilson, J. Q., & Kelling, G. L. (2006). A Quarter Century of Broken
Windows. American Interest, 2(1), 168-172.
Wolf, K. L., & Robbins, A. S. T. (2015). Metro Nature, Environmental Health, and
Economic Value. Environmental Health Perspectives, 123(5), 390-398.
Wolfe, M. K., & Mennis, J. (2012). Does vegetation encourage or suppress urban crime?
Evidence from Philadelphia, PA. Landscape and Urban Planning, 108(2012),
112-122.
47

Appendix A
The City of Olympia defined a downtown strategic area in order to implement a
comprehensive plan for increased safety, greater economic revenue, broader community
participation, and support for natural and historic spaces (City of Olympia, 2015).

Figure 8. City of Olympia Downtown Strategy Area.

48

Appendix B
During 2008 and 2015, the City conducted two LiDAR data collection events. The
following figures display the 2008 dataset’s 3.0 Foot Esri Grids for the LiDAR Bare
Earth and Highest Hit models that reference the lowest and highest laser return points,
respectively, and the 1.5 Foot GeoTiffs (high-intensity orthoimages) for both 2008 and
2015.

Figure 9. 2008 LiDAR Bare Earth return, 3.0 Foot Esri Grids, Horizontal Datum NAD83
(HARN), Vertical Datum NAVD88 (GEOID03).

49

Figure 10. 2008 LiDAR Highest Hit return, 3.0 Foot Esri Grids, Horizontal Datum
NAD83 (HARN), Vertical Datum NAVD88 (GEOID03).

50

Figure 11. 2008 high-intensity orthoimagery, 1.5 Foot GeoTiffs, Datum NAD1983. Inset
example demonstrates magnification capability for analysis.

51

Figure 12. 2015 high-intensity orthoimagery, 1.5 Foot GeoTiffs, Datum NAD1983. Inset
example demonstrates magnification capability for analysis.

52

Appendix C

Figure 13. Limitations due to inconsistent boundaries between census block groups and
patrol sector “B.”
53

Media

Loosle_WMESthesis2016.pdf