University of Southern California Dissertations and Theses

Mapping uniformity of park access using cadastral data within Network Analyst in Wake County, NC

(USC Thesis Other)

Mapping uniformity of park access using cadastral data within Network Analyst in Wake County, NC

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

MAPPING UNIFORMITY OF PARK ACCESS USING CADASTRAL DATA WITHIN
NETWORK ANALYST
IN WAKE COUNTY, NC

by

Jonathan Edward Parsons

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

May 2015

Copyright 2015 Jonathan Edward Parsons

ii

DEDICATION
I dedicate this document to my wife, Carrie, and my children, Silas and Matilda, for their
patience, understanding, and support throughout this entire process.

iii

ACKNOWLEDGMENTS
I will be forever grateful to my committee chair Professor Robert Vos who has kept me
challenged and pushed me to expand my research. Thank you also to my friends Yvonne,
Outhong, and Adrianna who have kept me motivated through the long nights to finish what I
started.

iv

TABLE OF CONTENTS
DEDICATION ii
ACKNOWLEDGMENTS iii
LIST OF TABLES vii
LIST OF EQUATIONS viii
LIST OF FIGURES ix
ABSTRACT xi
CHAPTER 1: INTRODUCTION 1
1.1 Defining Service Areas 2
1.2 Measuring Service Areas 5
1.2.1 Physical Access 6
1.2.2 Measuring Populations 8
1.2.3. Gap Analysis 9
1.3 Motivation 9
1.4 Case Study Wake County, NC 10
1.5 Organizational Framework 17
CHAPTER 2: RELATED WORK 18
2.1 Accessibility 19
2.1.1 Comparing Container and Coverage Approaches 21
2.1.2 Accounting for Physical Accessibility 25
2.1.3 Walkability and the Influence of Routes 30
2.2. Population Distribution and Dasymetric Mapping 31

v

2.3 Measurements of Park Availability 33
2.4 Limitations of Existing Park Accessibility Models 35
2.5 Summary 37
CHAPTER 3: METHODS AND DATA 38
3.1 Method Framework and Variables 39
3.1.1 Output Products 43
3.1.2 Data Sources 44
3.2 Key Variables and Formulas 46
3.2.1 Total Population per Parcel 46
3.2.2 Total Parkland per Parcel 49
3.2.2 Total Parkland per Person 49
3.3 Application of Network Analysis 50
3.3.1 Details of Network Analyst 52
3.4 Data Selection 53
3.4.1 Parcel Points 54
3.4.2 Park Data Preparation 56
3.4.3 Resolving Park Access Points 57
3.4.4 Systematic Approach to Locating Access Points 61
3.4.5 Preparing the Census Data 64
3.5 Dasymetric Mapping of Population Data 66
3.5.1 Final Output Table 67

vi

3.6 The Origin to Destination Matrix Tool Solution 67
3.7 Test Hardware and Software Setup 70
CHAPTER 4: RESULTS 72
4.1 Lessons Learned in Applying Network Analyst 72
4.2 Determining the Parcels Served 74
4.3 Output from the OD Matrix Tool 77
4.4 Comparison of Accessibility Measurements 88
CHAPTER 5: DISCUSSION AND CONCLUSIONS 96
5.1 Benefits of the Cadastral Based Measurement Technique 96
5.2 Disadvantages of the Cadastral-based Technique 97
5.3 Origin to Destination Matrix Tool Compared to the Service Area Calculation 98
5.4 Future Research 99
5.4.1 Adding Social Variables 101
5.4.2 Automation and Portability 102
5.4.3 Workflow Accessibility Evaluation 103
5.4.4 Scaling Workflow 104
5.5 Future Applications 106
5.6 Final Conclusions 107
REFERENCES 109

vii

LIST OF TABLES
Table 1 Source Data and Output Matrix 46
Table 2 Review of Residential Parcels within Wake County, NC 90
Table 3 Comparison of Selection Methods 91

viii

LIST OF EQUATIONS
Equation 1 Population per Household 47
Equation 2 Calculation for Total Population per Parcel 48
Equation 3 Total Parkland per Parcel 49
Equation 4 Parkland per Person 50
Equation 5 ODMT Output Check 68
Equation 6 Total Paths Possible ODMT 73

ix

LIST OF FIGURES
Figure 1 Contextual Map of Wake County, NC 12
Figure 2 Map of Parks within Wake County, NC 16
Figure 3 Example of Buffers Generated from Park Boundaries 23
Figure 4 Workflow Diagram Illustrating the Cadastral Based Uniformity Assessment 42
Figure 5 Model Showing Household Unit by Census Unit 48
Figure 6 Model for Isolating Residential Parcels and Joining Census Data 55
Figure 7 Residential Development Pattern of Wake County, NC 56
Figure 8 Model for Assigning Addresses to Non-addressed Parks 57
Figure 9 Map of Access Points at Bond Park 60
Figure 10 Defining Access for Linear Parks 62
Figure 11 Totaling Population by Census Unit 65
Figure 12 Illustration of Summary Data for Residential Parcels with Population Calculated 67
Figure 13 Service Area Map Based on Service Area Calculation 76
Figure 14 Raw Output from ODMT within ArcGIS 78
Figure 15 Output Table from the OCDM Calculation 79
Figure 16 Population Density per Parcel, Overall County 80
Figure 17 Population Density per Parcel, Focus View 81
Figure 18 Parks per Parcel, Countywide 83
Figure 19 Park Count within ¼-mile, Enlarged View 84
Figure 20 Acre of Parks within ¼ mile of Parcel, Countywide 86
Figure 21 Acre of Parks within ¼ mile of Parcel, Enlarged View 87
Figure 22 Summary Statistics for Number of Parks per Parcel for Wake County, NC 89

x

Figure 23 Histogram Graph of Park Acres per Parcel 91
Figure 24 Histogram Showing the Variation in Distance to Parks 92
Figure 25 Maps Showing Park Access Based on Nearest Distance to Parcel 94
Figure 26 Maps Showing Park Access Based on Nearest Distance to Parcel 95

xi

ABSTRACT
Park planners make long-term land acquisition and capital improvement plans based in part on
population growth and gap analysis of existing facilities. This study demonstrates a new
cadastral-based technique to measure park access for residents in Wake County, NC. Based on
road network and cadastral data, the technique uses the Origin-to-Destination Matrix Tool within
Esri’s Network Analyst extension in conjunction with dasymetric mapping of US Census Data to
the cadastral data. The demonstrated workflow provides for a highly detailed assessment of
walking distance between parcels and parks, that when linked with the population data, provides
a gap analysis based on the amount of parkland and number of parks available at each parcel.
Successful completion of an analysis at this level of detail illustrates a very different view of
park coverage for Wake County, NC compared to traditional methods, revealing how hard edges
created by major thoroughfares and soft edges created by property ownership impact pedestrian
accessibility. Using the cadastral-based method, 19.85% fewer parcels have 1/4-mile park
access than compared to a buffer based method (6.72% versus 26.27%). The use of this type of
technique will allow for a more comprehensive assessment of the peoples served by the park
system and when coupled with demographic information, may prove more effective in assessing
grants and monitoring the impact of public initiatives promoting equality and uniformity of
access to public parks.
1

CHAPTER 1: INTRODUCTION
The complexities of the urban environment often impede the ability of planners to predict
needs for and appropriately locate public facilities. In the case of public parks and open
space, understanding how pedestrians can access a park is necessary to evaluate the
effectiveness of the park system in serving the community (Nicholls 2001; Boone et al.
2009). In past studies, planners have evaluated the level of service by measuring the gaps
revealed between the parks based on buffers drawn around each park defined by a fixed
distance from a set point, what could be termed a coverage method (Nicholls 2001; Hass
2009). The distance often correlates to an assumed walking or travel distance. These
maps may not accurately identify the populations served, accurately depict existing
service gaps, or provide insightful planning strategies to improve park access.
The primary goal of this study is to develop a cadastral-based method for
measuring park access uniformity based on the actual travel routes. This technique
requires developing a workflow that solves three primary research objectives: 1) Develop
service areas based on walking networks; 2) Assess populations served based on physical
location; and 3) Calculate the acreage of parkland available to the population based on
the travel distance between the park and the parcel within the ¼ mile walking distance.
Developing an analysis tool that incorporates these objectives will provide planners and
community leadership a means to conduct repeatable analysis of park service in
residential areas.
Evaluating how this tool functions at a county-based scale will help determine its
ability to process larger study areas including potentially state wide and regional
assessments. Rather than measure the accessibility of any given park, the proposed

2

approach measures the access from the residential areas to all parks within a specified
distance, identifying residential areas that have gaps in service or have more parks than
the average. Grounding this process to accurate cadastral data will reduce the level of
generalization often involved in system wide analysis (Wolch et al. 2005; Maantay 2007).
The availability of a vetted Geographic Information System (GIS) to organize, manage,
and analyze complex data facilitates this type of detailed analysis. The study
demonstrates a GIS workflow that if widely implemented in local planning agencies
could lead to more effective uniformity and gap analysis of park access, identifying
where new facilities will be most effective.
1.1 Defining Service Areas
Service areas are determined by the distance a person is willing to travel to access a
facility (Hass 2009; Duncan 2012). How a service area is defined for a park will greatly
influence the calculations regarding populations served. Within the literature, three
distinct methods for measuring service area include container, coverage, and gravity
model techniques. The container method uses a social or political boundary to determine
the number of facilities within a given boundary, such as the municipal limits, and
compares the number of facilities to the number of people. The container method is often
used within larger scale geographic studies, such as multiple county or statewide
assessments, utilizing heavily aggregated data in order to identify spatial patterns and
trends at the risk of losing detail.
The coverage method utilizes buffering at a specific distance from a facility to
generate a polygon which is used in selecting which populations are potentially
influenced by the park based on how much of the underlying population data is covered

3

by the area of the park buffer. The gravity model, or similar nearest distance model,
assigns facilities to end users based on nearest distance. This approach is not typically
used because of the complexities in setting it up (Nicholls 2001); this study does not
explore the gravity model method as it is rarely utilized due to complexities in
establishing the criteria and calculations.
As described above, in the case of parks and open space, service areas are
typically measured using the coverage model and based on 1/4 mile buffers from the park
periphery (Nicholls 2001). The ¼ mile distance is an important variable for
consideration as it represents a round trip walking distance of ½ mile (Loukaitou-Sideris
and Stieglitz 2002; Sister et al. 2010; Hass 2009; Nicholls 2001; Talen 2010). The use of
¼ mile is an accepted standard based on the distance a family with small children is most
likely to walk to reach a park or service (Talen and Anselin 1998; Loukaitou-Sideris and
Stieglitz 2002). In areas of suburban development and a more car-centric development
pattern, the travel distance may increase due to the likelihood of parents driving to
facilities. In these cases, basing service areas on vehicular access may likely preclude the
use of the facility by families experiencing economic hardship (Loukaitou-Sideris and
Stieglitz 2002; Boone et al 2009).
Using buffers as a means to identify service areas can also mask social and
physical impediments that prevent accessing the park site. Physical impediments, such as
major transportation corridors or natural features such as streams and excessive changes
in topography may also be overlooked during a service area analysis using the buffering
process.

4

Both the container and coverage methods are susceptible to the influence of the
modifiable areal unit problem (MUAP). In the container method, it is the arbitrary nature
of the chosen boundary whether using objects like corporate boundaries or neighborhood
boundaries. Within the coverage method, generating buffers based on access points or
edge boundaries can create large discrepancies regarding final access statistics as not all
boundaries are permeable by the public. Similarly, using a buffer from the center of a
feature compared to the outer perimeter will generate differing coverage polygons and
skew the results in access assessment. Also, how the buffer intersects existing census
boundaries may change measurements. How the buffer is adjusted in terms of travel
distance also will lead to variations in results. In general, buffers generate an
indiscriminate boundary that does not reflect the actual condition of the user group. A
person just outside of the ¼ mile buffer may be just as likely to use the park if there is a
safe pedestrian path as someone just inside the ¼ mile is unlikely to use the park if there
is no such path.
Parcels that are intersected by the buffer may not be able to access the park based
on road networks. Using service areas generated from network analysis will create
service areas that reflect the physical access paths. This approach will allow repeatable
results that can be modified based on the mode of travel, size of the park and, in the
future incorporate a form of costing. This study will utilize such network-based service
areas to assign and validate populations to the respective parks, assign park area to the
respective parcels, and allow for charting of both work and market development.
One advantage of a gravity based model is that it considers how a park service
area may increase based on the park type. Park facilities vary in the level of services

5

provided, whether they are a large regional park or a small neighborhood style park. An
example of a large regional park would be Pullen Park in Raleigh, NC. This large
regional park provides community residents access to an aquatic center, carousel,
playgrounds, and even paddle boats. Compare this to a small neighborhood park, such as
Jaycee Park in Raleigh, NC, where the amenities are less diverse and focus on more
traditional offerings such as a playground, basketball court, and picnic shelter. The
critical contrast in these two park types is the distance that people are willing to travel to
access the amenities. People may travel from as far as Cary, NC at 12 miles, or Wake
Forest, NC at 14 miles to use at Pullen Park, while they would remain in their own
communities when using playground or picnic shelter. This variability in amenities
influences the draw power of specific parks when determining what distance people are
willing to travel and influences how to measure service areas (Loukaitou-Sideris and
Stieglitz 2002). Although this study focuses on basic pedestrian access for any type of
park, the overall method of measurement presented here is not inconsistent with a
gravity-based model.
1.2 Measuring Service Areas
The generally accepted approach for assessing park service areas is generating
buffers representing service areas from park features, what is termed the coverage
method in this study (Sister et al. 2010; Boone et al. 2009; Hass 2009). As mentioned in
the previous section, the buffer approach, or coverage method, uses a fixed distance
buffer from the edge of the park property to generate a polygon that represents the service
area. The size of the polygon will depend on the variables used during the buffering to
represent the assumed travel distance, whether walking or driving. This polygon,

6

representing the service area, aids in a number of assessments including calculating
populations served, assessing the system for gaps in coverage, and identifying sites for
future parks. Based on findings from the previous studies, this study hypothesizes three
main sources of error to the traditional service area measurement approaches: 1)
Generalization of physical access to the park, 2) Spatial inconsistencies in assessing
populations served, and 3) Gap analysis in coverage that does not represent real use
patterns.
1.2.1 Physical Access
Physical access is a primary concern when determining the service area of a park. The
common utilization of container and coverage-type approaches often overlooks the
nuances of physical access due to either the scale of assessment or lack of detailed
information. Coverages generated from buffers are generalizations, overlooking elements
such as major roadways, railroad crossings, and property boundaries (e.g., fences) when
they are generated using a fixed radius from the park edge.
How a person perceives these varying objects in terms of boundaries, or barriers,
involves understanding the differences between the hard and soft barriers that compose a
complex urban environment (Lynch 1960). Hard edges, defined by the physical
obstructions as noted above can be readily identified and can be measured in how it
influences how people may travel through an area; consider something as limiting as a
river or a set of high speed railroad tracks. Hard edges are often already recorded in
geospatial datasets and thus easy to account in GIS network analysis (Makse et al. 1995).
Soft edges are harder to grasp within GIS. The concept of a soft edge is grounded
in how people move in and construct a mental image of a city or urban area (Lynch

7

1960). The urban fabric in its pure form is one of permeability, allowing people to move
freely through the grid, with no hard physical barriers separating them. The reality is that
how people perceive the built environment, in terms of built character, social makeup,
and physical location in relationship to the overall context serve to create soft edges
within a person’s mind. It is often related to perception on how people feel about the
environment they are within and whether they can traverse across a property or
communal path to get to a different portion of the environment.
An example of this soft edge phenomena occurs when you observe the emotional
reactions as people cross from one area of the environment to another of differing cultural
values, such as crossing from a residential area to an industrial area, or from a poor
residential area to an affluent residential area. Due to the variance in reactions on a
person-by-person level, it would be difficult to map the varying levels of permeability
within an environment and how it can restrict or allow people to travel through the
community without extensive human subjects research (Lynch 1960). However, certain
assumptions for this study, for example, include people being unwilling to traverse
private property boundaries like their neighbors backyards to enter a park.
An improved method for measuring service areas will need to account for
physical access routes that connect the parks to the populations served. As these
pathways are often well-defined and perceived as public in ownership and access, they
are the most easily mapped within a GIS. Understanding where park access points are in
relationship to the potential users will be key in developing accurate service area
measurements.

8

How the access points interact with the adjacent vehicular and pedestrian routes
may also have a significant influence over how users can access the park. Developing a
model that has the ability to adjust the travel distance to match the mode of travel will
provide analysts flexibility in determining the service areas. The use of actual access
routes will also aid the analysis by excluding properties that are not within the travel
distance along the accessible routes, and even potentially offering opportunities for
including barriers in the analysis. Details regarding of how these nuances integrate into
the approach are described within Chapter 3 of this paper below.
1.2.2 Measuring Populations
A primary goal of developing a service area measurement is to assess what populations
are served by a particular park facility. The primary source for population data in park
access studies is U.S. Census data. Census data contains population data in one of several
different scales depending on the target application. Using the smallest scale of measure,
the census block group, analyses are often conducted where the population is calculated
by the percentage of the block group within the service area. This poses an interesting
challenge when service areas are general buffers. Populations that are within the ¼ mile
buffer radius may not necessarily be within ¼ mile travel distance. Using physically
grounded service areas in conjunction with accurately located housing unit information
may offer significant improvements in estimating park accessibility.
Dasymetric mapping is a form of thematic mapping that assigns data based on one
areal unit to another (Maantay 2007). In the case of population analysis, the ability to
map the census population data by a shared attribute to the parcel data will allow the
population to be mapped to the actual residential areas. This may prove a more accurate

9

means when evaluating the accessibility of parks as only those units within the service
area will be selected. Furthermore, this reassignment of population information will
allow a ranking analysis to be completed based on the amount of parkland available per
person at a parcel level.
1.2.3. Gap Analysis
A gap analysis represents the culmination of measuring the individual factors and
assessing the effectiveness of the park system. A gap analysis looks at the spatial
arrangement of the parks within a given system in relationship to the people they serve
based on the park service areas. The output of the gap analysis will identify where there
is no coverage by parks. For this study, the intended direction of the gap analysis is to
reverse the direction of measurement from traditional approaches by starting at the parcel
and searching outward for connected parks rather than vice versa.
This concept of a cadastral-based gap analysis is focused on understanding the
uniformity of coverage within the park system. By measuring the quantity and area of
park available at each parcel within the specified travel distance, the resulting gap
analysis maps will help park planners understand how the existing park system serves the
community and show where the concentrations of park accessible residential parcels are
and where shortfalls in coverage exist. To be clear, this type of measurement will quickly
identify those parcels that have more or less park available compared across the study
area based on the physical access available to that parcel.
1.3 Motivation
The motivation for this project is rooted in the need for an analysis tool that allows park
planners, landscape architects, and municipal leaders to quickly understand the level of

10

service of the park system based on the real world urban fabric. Development pressures
often outpace the acquisition and development of new park sites. Creating an approach
that is tied to cadastral data would allow the planning professionals to have an up-to-date
view of where pressures in the system will occur, harnessing the information collected
including changes in land use and projected new housing areas based on housing unit
counts. It opens up the opportunity for predictive modeling as a means vetting park sites.
Current practices of finding the cheapest land available without checking the physical
access creates parks that are underutilized and is a misallocation of scarce community
resources.
If combined with other datasets, the method demonstrated here could also be
used to quantify and map the population of children under the age of 18 as a means of
more accurately assessing park locations against the population most likely to utilize the
park facilities (Maroko et al. 2009). This is in contrast to comparable studies that have
focused on the entire population base as a metric. Healthy lifestyles begin during the
early childhood years, understanding the access children have to public parkland can be
useful as an evaluation criteria. This method may also reveal where families are
concentrated compared to other household types, which may allow for assessing where
funds are needed for additional play equipment or for adult activity types such as tennis
or fitness trails, providing insight on locations of need (Dalton et al. 2013).
1.4 Case Study Wake County, NC
The focus of this study is to assess the accessibility of public parkland within the
distributed development patterns and resultant population centers in Wake County, North
Carolina. Wake County is located in central North Carolina and covers 835 square miles.

11

The county encompasses Raleigh, the state capital, as well as eleven other incorporated
municipalities. As of the 2010 census, the county population was estimated at 900,993
people. Figure 1 shows the contextual relationship of Wake County, NC within the
southeastern region of the United States.

12

Wake County is fast growing in both population and housing. The total
population of Wake County grew over 44% between 2000 and 2010 (Wake County
Figure 1 Contextual Map of Wake County, NC

13

2013). This growth establishes Wake County as the second most populous county in the
state. A 2012 estimate calculates the population of Wake County at 952,151 people with
over 181,255 under the age of 15 (OSBM 2014). The North Carolina State Demographer
projects that the population will exceed one million residents by the year 2015.
The fast growing population has increased demand for new housing, amenities,
and recreation areas. The City of Raleigh estimated an addition of 55,425 housing units
for an increase of 46% in units between the 2000 and 2010 census (Raleigh 2014). The
expansion of housing places strong demand on the remaining undeveloped land within
Wake County and may push populations away from Raleigh. As this development
advances from the urban core, the pressure on existing parks and the need for new parks
may be pronounced.
Early subdivisions in Wake County were not required to provide public open
space and parks at the time of development. In planning, the concepts of open space and
parkland are often interchangeable. However, to be more specific, open space is a broad
term that can include natural areas, state parks, forests, wildlife preserves, and even
preserved agricultural lands. In contrast, for this study, parks are areas that have facilities
for rest and recreation owned and managed by the city or county for the enjoyment of the
public. The intent of both parks and open space is to provide balance between built land
and undeveloped lands. While Wake County and the individual municipalities have
strived to provide open space and parkland for their respective populations,
disconnections exist in the quality of the individual facilities in terms of physical size and
amenities. These disconnections can lead to user groups utilizing facilities in adjacent

14

areas, overburdening those facilities, or avoiding using parks due to lack of convenience
(Dalton et al. 2013).
Parks are a significant amenity as they provide an environment that encourages
increased physical activity through active recreation and play (Dalton et al. 2013). As
development increases, parks also aid in preserving and improving the environmental
quality in dense urban developments (Carleyolsen et al. 2005). Ensuring that park
facilities are equitably distributed in terms of geography and accessibility is a
requirement for the success of an existing parks system. Accurately mapping the
relationship between facilities and population is key to identifying gaps in facility
coverage (Duncan et al. 2012).
It is also important to ensure that these facilities are equitably distributed in terms
of need. Historically, park need assessments focus on dense urban areas that have a
higher demand for parkland to meet recreational opportunities when compared to
traditional suburban development (Wolch et al. 2005; Sister et al. 2010). In Wake
County, development has spread from the central core of Raleigh through multiple
independent subdivisions, often referred to as urban sprawl. This sprawl has resulted in a
decentralization of development in Raleigh, NC. The only dense urban development in
Wake County aligns to core area of Raleigh, while the remainder of Wake County is
relatively low-density, single-family housing. The average density has the zoning
classification of R20 or one residential unit per 20,000 square feet of yard area (Wake
2014). A low-density development pattern is difficult to adequately serve with public
facilities.

15

Another challenge facing park planners is the notion that single family homes
provide large yards for children to play in. This is based on the traditional suburban style
of detached houses on large green lots (Marusic 2011). As the demand for housing and
the cost of land increases, lot sizes are being reduced to increase density. This limits the
amount of private green space available to the individual subdivisions. The challenge to
this model is the need for children to have varied experiences for play and socialization
for healthy development (Loukaitou-Sideris et al. 2002; Sister et al. 2010).
With the increasing rate of residential development on the remaining vacant land
within Wake County, the location of the existing parks and their respective service areas
is important in evaluating the effectiveness of the park system and its ability to continue
to serve the increasing population. Risk exists in the population spreading beyond the
reach of existing facilities. Without proper planning, there will not be adequate land for
park construction within these newly developed areas due to higher land costs
(Carleyolsen et al. 2005). As areas increase in density, existing green areas will continue
to be reduced placing a higher burden on existing parklands.
In the denser residential areas of Raleigh, parks are integrated into the
neighborhoods. Residential areas close to the center of Raleigh maintain high land values
that preclude young and lower income families from living within these more established
neighborhoods. As the cost of living continues to drive the population further out from
the core, the risk of isolation from public parks will increase due to travel costs, access,
and geographic location. Figure 2 shows the number of parks within Wake County, NC.
At the time of this study there are 268 distinct parks defined within the attribute data of
the Wake County Open Space inventory data file (WakeC 2014).

16

Figure 2 Map of Parks within Wake County, NC

17

1.5 Organizational Framework
This thesis is organized into four additional chapters. Chapter 2 focuses on
reviewing the literature available and utilizing these examples to identify and discuss the
types of service assessments and population measurement techniques presently in use to
analyze park accessibility. A discussion of benefits and constraints observed from these
studies will also serve as a foundation for the proposed methodology. Chapter 3 focuses
on presenting a framework for the proposed cadastral-based measurement technique.
This chapter will describe the framework necessary to develop a tool that integrates
service area generation through network analysis with dasymetric mapping of
populations. The process for developing the application and operational concerns is
discussed within this section. Chapter 4 will review the results of the proposed modeling
technique. This chapter will present the model outcomes, discuss the benefits, outline the
shortfalls, and identify improvements necessary. Chapter 5 will discuss the implications
of using this revised approach to generate gap analysis and identify opportunities for
future research and exploration. Future research and explorations will also be described
within this concluding section.
18

CHAPTER 2: RELATED WORK
The realities of human nature lead us to deduce that people will only use parks that they
can conveniently access whether by car or by foot. This simple premise defines this
study. In order to effectively assess a park system for uniformity of access by the
community’s residential areas, a measurement technique needs to be developed that is
parcel centric. A parcel centric approach focuses on utilizing the cadastral data and travel
routes to determine which parcels have access to parks. This approach will provide
analysts a means for comparing the quantity of parks and area of parkland available to
each parcel and its residents. A common benchmark is to compare the capacity of a park
system on the notion of acres per 1000 people available (Moeller 2014). Thus, an
effective gap analysis needs to be based on an underlying acres per capita measurement
(Hess 2001; Cary 2012). This will provide analysts a clear view of where gaps exist in
terms of level of service and inequality in terms of the populations served.
The purpose of this chapter is to explore the existing literature and identify the
issues surrounding park accessibility in terms of determining service coverage to the
resident population. How have previous studies evaluated physical access and measured
uniformity in terms of park space and amenities? Attention is paid to refining the
definition of accessibility in regards to the literature and how it will function within this
study. Defining accessibility will also require reviewing common measurement
techniques. The literature review will also discuss the issue of utilizing a park-centric
approach in the current methodology and provide practical examples. This assessment
will establish the groundwork for presenting a new measurement approach that focuses
on the parcel as a means to evaluate overall park accessibility.

19

To facilitate this discussion, the chapter is organized into four main subsections:
accessibility, dasymetric mapping of residential populations, equity analysis, and a
review of practical examples. Through understanding of these variables and how they are
influenced by the existing techniques, a foundation for developing a new approach for
measuring park accessibility can be laid. This foundation will support the need for a
parcel centric approach in lieu of the park centric model currently employed. After the
main points are discussed in terms of process and execution. At the conclusion of this
chapter, a summary will be provided that outlines how these findings inform the method
proposed and demonstrated in this study.
2.1 Accessibility
Accessibility can have broad meanings. For this study, accessibility will relate to a
person’s ability to physically reach a specific park based on known travel routes. This is a
more focused view than what is common in the literature, where accessibility is often a
combined measurement of equitable distribution determined by both physical and social
dimensions (Lindsey et al 2001). Physical dimensions are defined as those tied to
coverage-based analysis and travel distances and social dimensions are defined as
relating to the perception of a user to traverse a perceived soft edge to reach the park.
Within these two vectors, there is often a third measure, which ties back to the area of
park available to the people served (Higgs et al. 2012). Focusing on either attribute
without respect to the other will result in only a partial assessment of a park system’s
ability to serve a given community. In the simplest of forms, accessibility of a park
system may better be represented as a measure of total opportunities per person to a given
origin within a given distance (Hass 2009). Following from this definition, there are

20

typically three primary steps necessary for collecting the information necessary to
evaluate accessibility and they are as follows: determine the study area, establish the areal
selection unit and analysis method, and conduct the calculations based on the areal unit.
Through the review of existing literature, there are two primary categories of
technique deployed: the coverage method and the container method (Tarrant and Cordell
1999). Both methods can determine the potential populations served and aid to complete
the final analysis regarding demographics and equality assessments. The following
subsections contain a detailed comparison of how these two processes work within an
analysis workflow. The focus of this study is to demonstrate whether a cadastral-based
approach can replace the traditional methods.
Several studies have focused on the aspect of the total measure of recreational
opportunities as a means to determine the equality of park distribution (Talen & Anselin
1998; Lindsey 2001; Sister 2012). Access to parks has been found to relate directly to
the health of a community (Carleyolsen et al. 2005). Developing a measurement
technique that can respond to the variety of facilities within a given distance of a
residence remains the issue when trying to evaluate the availability of specialized park
facilities within the greater community. Locating facilities such as aquatics centers and
fitness centers in areas that are accessible by the greatest number of residents is key to
maintaining a uniform park system. It is also an important aspect to account for the
number of people served when setting aside space for parks or open space and for park
planners when considering resource management and capital improvements for park
amenities (e.g., pools, tennis courts, and playing fields).

21

2.1.1 Comparing Container and Coverage Approaches
The general process for a container analysis is to identify the region of interest, determine
what boundary or area the calculations will use as the unit of analysis. In most studies,
the container approach is based on a political boundary such as county or city jurisdiction
and serves as a selection tool to select the population data and parks within the given area
in order to complete calculations revealing the amount of parkland available per person
(Talen & Anselin 1998; Tarrant 1999; Wolch et al 2009). Using the container approach,
input data is selected based on how it intersects with the overlaying study boundary.
Regarding parks, the total area of all parks within this unit of analysis is calculated. Once
complete, this area is divided by the total population of the unit-of-analysis. In studies
focused on quantity of parks, the area-person-calculation is replaced by a count of parks
available within the unit of analysis. One issue revealed in this approach is how one
handles the calculation when only part of the population feature or park feature is located
within the container. Determining whether to use the percentage of intersection, or to
exclude these intersecting data, can greatly influence the output results.
A key limitation of the coverage approach is that it uses arbitrary boundaries, such
as political jurisdictions or census tracts, which contribute to the Modifiable Areal Unit
Problem (MAUP). MAUP is an issue in spatial analysis that is created by generalizations
that may be introduced into GIS data. The generalizations, whether based on establishing
arbitrary boundaries such as political jurisdictions, or by an aggregation of the spatial
phenomena being mapped such that the boundaries may not match with the desired scale
of the analysis. Census data is susceptible to the MAUP as the boundaries change each
census depending on the rate of growth within a specific area and the intent of

22

maintaining each census unit at a specified range of population and households. In most
instances, this is a maximum of about 3500 people. Figure 3 provides an example of
buffers generated from the park edge at a ¼ mile fixed distance and the access roads that
either front the parks or directly access the parks. This exhibit shows the nuances
associated between physical access and the arbitrary service areas drawn by buffering of
the park. These roads serve as the potential points of access when examining the
boundaries of the surrounding properties. Looking at Davis Drive Park, the eastern edge
borders a railroad, the south and north by private property, all three of which limit the
access to the west side. Fred G. Bond Metro Park is an extreme example showing the
amount of park bordered by private land in comparison to road frontage.

23

The coverage approach is seen in studies like the LA Green project (Sister et. al
2011). In the LA Green project, the park distribution of Los Angeles County was
assessed utilizing the census tract data as the boundaries to evaluate the parks available to
those populations of each census tract. Using this approach creates a focused view of
parkland by census tract that may or may not reflect how the population circulates within
the urban fabric. However, such an approach allows analysts to compare the population
data to park data with relatively minor interaction within the GIS environment. This is an
Figure 3 Example of Buffers Generated from Park
Boundaries

24

important consideration when working with large-scale study areas where detailed
analysis may either be computationally prohibitive or the differences in results may be
statistically insignificant given the study inputs.
In comparison, the container approach starts from the parks themselves and
generates a buffer that serves as an overlay for use in selecting the populations or
locations influenced by a given park. In most cases GIS is utilized to generate a fixed
distance from either a center point or edge of the park feature to generate a polygon
shape. This creates a polygon that is then used to select or exclude data as necessary from
the analysis. Nicholls credits the initial coverage workflow to R.L. Hodgart (Nicholls
2001). This is clearly a park centric approach.
The container method often claims to generate service areas for parks. A service
area, as a simplified definition, is as a means to generalize the travel distance from the
park out into the community into an area that will likely draw users. This is useful in
potentially identifying how many of a specific population may use the park (Nicholls
2001; Talen 1999). The typical approach utilized to generate a service area is to create a
buffer around the feature of interest, in the case of this study a park. The buffer is
generated either from the centroid or from the perimeter of the subject property. The
buffer distance is based on the mode of travel being tested, such as walking or driving.
(Talen and Anselin 1998; Nicholls 2001; Hass; 2009; Joseph 2011)
To measure the population within the resulting service area, the analyst can then
select features based on location, selecting those underlying objects that are within the
cover of the service area. For those objects that either are at the periphery or partially cut
off, the analyst can calculate the percentage of the population within the service area

25

based on the area of overlap (Boone et al 2009; Nicholls 2001; Tarrant 1999). The
coverage approach, when using shapes generated by buffers, is often selected because of
the approximation that can be completed for reflecting travel patterns of those
populations that may use the park. This approach also follows the park planning
guidelines as it allows planners to easily create coverage circles based on park type and
size (Talen 2010).
The challenges facing a coverage based on buffers are rooted in the risks of
generalization of the ground condition. Whenever a buffer is used there is a risk of
generalization whether from spatial estimation or oversimplification of the ground
features. The precise dimensions of the buffer is also a factor of whether the analyst uses
a centroid or boundary object (Nicholls 2001; Hass 2011). This falls within the realm of
the modifiable areal unit problem (MAUP) which is a fundamental source of error when
conducting GIS Analysis (Talen & Anselin 1998). This will be covered in further detail
within Chapter 4 through a comparison of parcels served as selected through a coverage
analysis based on buffers and a selection set based on the Origin to Destination Matrix
Tool.
2.1.2 Accounting for Physical Accessibility
Planning doctrine tells us that parks need to be located where they will serve the most
people. This typically translates to those sites that can be easily reached by walking or
driving (Talen 1997). The influence of physical access on a populations’ ability to use a
park is a key variable to accurately measure in park accessibility assessments. People will
only travel to a destination that provides tangible benefits that exceed the effort in
reaching the destination (Tarrant 1999 & Cordell 1999; Nicholls 2001; Hass 2009). This

26

potential is akin to the idea of gravitational pull: the amount of distance users are willing
to travel relates to the size and type of park. Larger parks with a larger number of
specialized amenities will pull populations from further away. This is noted in planning
doctrine regarding appropriate planning metrics for park quantities based on park type
and population rates. For example, the Town of Cary located within Wake County,
North Carolina reflects this phenomenon in their use-ability assessment by assigning
differing travel distances between community level parks and their metro or regional
class park (Cary 2012).
It is possible to analyze the actual transportation routes that the target population
may utilize to access a given park using a GIS. This physical accessibility is determined
using digitized road and park data. The benefit of utilizing GIS to conduct a physical
accessibility analysis depends on the availability of the data. While a more detailed
overview of the process will be covered in the methods section, the use of network
analysis provides a means to calculate service areas maps based on potential travel paths.
While no one case represented a complete accessibility assessment in which population
data and park statistics are integrated, there are several key points worth discussing in
regards to success and potential for error.
In the work done by Nicholls (2001), network analysis served as a means to
generate the service area to conduct a traditional coverage based analysis. This approach
utilizes road networks to develop the service area for each park based on the access
points as mapped by the field team (Nicholls 2001). The roadways approximated the
sidewalks and normal paths of travel used by park patrons. Greenways and other
footpaths were not included in the study, nor were large regional parks or community

27

centers where vehicles are the primary mode of travel. Parks with multiple access points
and derived service areas combined into single service areas using the union tool within
ArcGIS. This creates one simplified service area per park for analyzing coverage. This
approach is similar to work done by Hass (2009) in testing of the use of Service Area
Calculations within ArcGIS Network Analyst to determine regional park access.
Within the Nicholls (2001) study, Network Analyst derived service areas were
compared to buffer-generated service areas. In both applications, the respective service
areas were generated using a ¼ mile as the distance. The buffer was generated from the
boundaries of the parks. The resulting differences between the parks were compared by
subtracting the difference in areas between the two coverages as well as sampling the
differences in populations. This analysis was conducted using simple, two variable
statistical comparisons. This work was done completely within ArcGIS as sampled using
the “select by location” feature within arc GIS.
Comparing the two methods of coverage generation revealed appreciable
differences between the network analysis and the buffer based coverages (Nicholls 2001).
Nicholls (2001) concluded that the largest variable in the difference in area definition was
the use of road network to derive the coverage. This created smaller areas due to the
types of road networks present within the study area
Nicholls (2001) carried out his study in Bryan Texas. Testing these differences
further, a comparison was made between the population sampled from the buffer model
and from the traditional coverage model. The difference was calculated at over 8,547
people, a significant difference when considering the accessibility of the people to the

28

population. When comparing just small neighborhood parks, the difference between
service techniques was much smaller, less than 7% (Nicholls 2001).
In both conditions, it is important to consider issues with sampling the population.
The census tracts were selected based on the location within the coverage of the service
area. This created the potential for errors regarding arbitrary census boundaries. This
opens the analysis up to the Modifiable Areal Unit Problem (MAUP), as portions of the
census block units that are within the coverage may or may not be inhabited, yet still be
selected and thus included in the population calculation (Holt and Hodler 2004). This
can lead to miscalculations in the percent in coverage of the population for those people
within the service area in both accords. To correct this, parcel data or physical household
locations needs to be utilized to better identify how many people are within the service
area.
A secondary issue of this network-based analysis involved the classifications of
the road system within the network analysis. The hierarchy of the roads as stored within
the Wake County streets dataset allows classifying thoroughfares that are not conducive
of pedestrian circulation, and as residential collector class streets, which are pedestrian,
oriented. The road network was not physically assessed to determine which roads were
or were not passable for pedestrian traffic, as this would require significant resources
beyond the scope of this study. This study uses the road generalization to test practical
application, a more detailed road network dataset would be needed to account for
assessing the nuances of road design that influence pedestrian circulation in the future
analysis (Nicholls 2001).

29

From the Nicholls study, there is a noticeable gap in application of network
analysis in determining park equality. The logic revealed in the work done by Nicholls
(2001) shows that it is a viable method to consider when detail is necessary for evaluating
a park’s physical accessibility when compared to the straight line buffer approach (Liu et
al. 2014). It also shows that there is further work needed in terms of understanding the
relationship of physical and assumed barriers within a community in regards to access
and how these may influence the measurement technique.
The most direct influences on access are physical barriers, such as large
transportation thoroughfares, railways, and natural features such as rivers and
topographic barriers (Lynch 1960). The physical barriers are often easier to define and
account for during an analysis. Soft boundaries, such as political or social boundaries,
are often more difficult (Lynch 1960; Nicholls 2001; Cary 2012). Network analysis can
incorporate a variety of barrier features to reflect on costs or increased weights in order to
access a specific park (Esri 2014).
This study accounts for the physical barriers through the classification of the
roads; by limiting the road types to neighborhood collectors and residential streets, unsafe
pedestrian routes associated with highways and thoroughfares should be excluded. It is
also assumed that the road network already facilitates the necessary crossings over water
bodies and topographic barriers. How the soft barriers may influence the likely
population will need to be explored further. These concerns will be discussed within
chapter 5.

30

The drawback in both the Nicholls (2001) study and a similar network approach
tested by Hass (2009) is how the population is measured within the service area district.
In both accounts, it is coverage based sampling of the population data. The strength of a
route based analysis is the ability to select those addresses, or properties, along a given
route that are able to access a given park facility at the distance specified. This is
important to consider as an improvement to the Nicholls (2001) and Hass (2009) network
analysis techniques.
2.1.3 Walkability and the Influence of Routes
Walkability as a concept is important for determining physical accessibility. The
likelihood of a person to utilize a park is dependent on the convenience of accessing that
park in terms of cost, or resources necessary, in comparison to the opportunities that it
offers (Talen and Anselin 1998). An accepted planning metric is to utilize a walking
distance between 0-¼ miles with the total round trip being less than ½ mile for those with
small children (Nicholls 2001). The ¼ mile is tied to the National Recreation and Parks
Association standards for the distribution of neighborhood parks to be within ¼ mile or
less of residents being served and not blocked by arterial roads or other physical barriers.
The accepted standards for park area per person is established in research conducted by
the American Planning Association and the NRPA, and assesses park coverage on a
acreage per person unit (Moeller 2014)
In economic terms, this means the benefit of experience offered by the park must
exceed the cost, whether time, energy, or actual capital necessary to reach it. This
concept of cost is further complicated by the nature of human activity. In an urban
environment, the likelihood of a person to utilize a park is based on its location along the

31

main transportation routes that that person may use on a regular basis. An illustrative
example to consider would be how a person travels along an imaginary Main Street to
reach a grocery store from the primary residence. A park that is located along Main
Street between the house and store is more likely to be utilized than a park that is located
beyond the store on the same path. A park located off to the side of the main travel path
along a side road would be even less likely to be utilized as the costs to access increases.
This is where coverage based analysis breaks down in assessing the potential user-base of
a park. While this study is not focused on economic modeling of the transportation
network, the importance of understanding the road hierarchy regarding travel paths and
how this hierarchy can influence the user group is important whether specific marginal
costs and benefits are estimated.
2.2. Population Distribution and Dasymetric Mapping
The physical location of residential properties to the park can be overlooked through the
generalization of the population data when conducting gap analysis. Current practice has
focused on the use of census data aggregated at the level of the census tract or the census
block group as the primary means to calculate the people served by a particular park
(Tarrant 1999). While this method produces practical estimates, the increasing quality
and detail of GIS data warrants the exploration of improving this process to reflect the
ground conditions more accurately.
Census Data is inherently a generalization of where people are located within a
given community. Even at the smallest scale, or most detailed unit, the census block
group, there is a large amount of generalization in regards to actual household locations
within the areal unit.

32

While the differing units of scale can improve the detail of the areal aggregation,
the units themselves are still generalizations of the ground plane. This means that
locations of houses, apartments, and other residential structures are not specifically
known. When doing a distance-based analysis to determine accessibility, population
sampled based on area and not on household locations creates concern, specifically in
situations that focus on the use of coverage methods.
Such detail may not be necessary at larger regional scale assessments, where a
container-based analysis is being utilized to do rough order magnitude assessments.
However, there is need for a higher level of detail when considering park sites. The use
of a parcel-centric analysis approach is one way to provide such detail. The work by
Nicholls starts to set the stage for developing a service area mapping method that
mitigates the hard cut-off but continues to use the census data units rather than applying
the same level of detail to refining the population mapping through dasymetric mapping
(Nicholls 2001).
A means to resolve this is by mapping the population to the ground condition
through re-association, or dasymetric mapping. Dasymetric mapping is a process in
which an analyst can reassign demographic attributes typically from Census data from
one areal unit to others based on determinations from alternative map layers. In the case
of population, with appropriate cadastral data layers, it is possible to reassign the
population based on the total residential tracts within specific census unit using either lot
area or housing units as means to determine the ratios.
Maantay et al. (2007) proposed a technique for modeling population density
against the urban environment using the parcel data to provide density that is more

33

realistic and location based maps. The system, referred to as the Cadastral-based Expert
Dasymetric System (CEDS), projects the census population information onto the target
parcels, aggregating the population based on the size of the parcel divided against the
total area of all of the parcels within a given census unit. Within the CERD, where the
parcel intersects multiple census areas, a calculation is conducted to add the population
proportional to the coverage. This approach works within existing GIS data as opposed
to the raster based approach proposed by Langford (2007), in which the underlying
subject area is rendered at a value per pixel rate, with which the total is then calculated by
means of sampling the pixels within a given selection area.
In this study, a similar concept is proposed to Maantay (200, but instead of doing
a strict areal conversion, it will be possible to assign population based on the household
units per parcel. Wake County GIS maintains an active database for all of the property
within Wake County with attributes that depict residential uses, structure types, and total
number of household units on a given parcel (WakeGIS 2014).
2.3 Measurements of Park Availability
The measurement of park availability is divergent in the literature. Depending on the
source, availability is either measured by the number of parks available within a given
service area (container method) or the amount of parkland available per person is
measured in regards to the geographic area covered (coverage method). This creates a
challenge when determining the most appropriate metric to utilize.
There are benefits and drawbacks to each approach. The initial response may be
to provide a tool that calculates both as that will ensure flexibility for the analyst in terms
of final output.

34

Within the planning community the guidelines for park planning is to provide a
minimum of 10 acres of open space per 1000 residents square feet per person as noted by
the National Parks and Recreation Association (Nicholls 2001). This measurement is
assumed to provide the diversity necessary to encourage active play and recreation when
influencing a healthy lifestyle (Lindsey et al. 2001). This standard is also compounded
by the need to provide a diversity of facilities within a given system (Moeller 2014).
However when there is only one park available within a given distance, this may have a
negative influence on diversity (Sister et al. 2007). Appropriate methods for determining
the area per person available require developing a means to isolate those people within
the park service area accurately.
In contrast to the area-per-person metric, some metrics on the number of parks
available to a given community (Moeller 2014). By counting the number of parks within
a given community, an analyst has a means to evaluate the potential diversity of the park
system within that community. This is an important consideration as in a community.
With one single large park, there are benefits to the larger space. However, there is only
one facility, whereas, by including neighborhood parks of varying size and character,
communities may offer users a more diverse experience.
Both aspects of park measurement must be considered when assessing a park
system. From a planning perspective, having numerous small parks in a specific area will
show a high level of accessibility if utilizing a count based criteria. However, using a
count based criteria in conjunction with an area per person based counting would allow
for a park system to be evaluated more effectively since the system is being evaluated for
both the potential diversity of location as well as overall quantity of space available.

35

How these two data points are stored within the final database can help lead to further
analysis when elements such as park facility types and quality of these facilities
integrated into an overall database.
Social access is a measurement of the diversity of people served by a specific
park. This type of access would include measuring factors such as age, ethnicity, and
economics. This study is focused on physical access and how parcels access parks;
exploring the distance and uniformity of this relationship. Determining the mode of
travel guides the direction of the analysis in regards to search area distance from a park.
This form of areal analysis defines the service area of a park. The mode of travel,
whether by bike, car, or walking, can also influence which residents are able to access the
park (Tarrant 1999; Heckert 2013). How accessibility is measured will have great
influence over the planning of the park system.
2.4 Limitations of Existing Park Accessibility Models
Within GIS, the creation of digital representations of spatial phenomena can create
generalized and even arbitrary representations of real world conditions. The coverage-
based analysis method in itself is an arbitrary means of analysis often based on political
or social boundaries. These boundaries based on geographic features, roads, assumed
break lines between changing values, or any combination of other factors. Both the
coverage and container methods are susceptible to the Modifiable Unit Area Problem
(MAUP) as they create a generalized if not arbitrary approach to determining the access
to the park.
Physical access routes will not always correspond to container analysis and
coverage based service area measurements. Accessibility studies often assume that all

36

sides of the facility are open and clear, with access available from all of sides. This
accessibility assumption is not always true due to physical constraints on the ground that
prevent ready access to the park facility along sidewalks and public rights of way
(Moroko 2009). In an extreme example, research testing the accessibility of users to a
specific store found that while the buffer map distance measured 4,500 feet, the actual
travel distance approaches three and a half miles (Nicholls 2001; Clift 1994). The other
challenge of the buffer approach is that it creates a sharp cutoff instead of a decaying
parameter.
The concept of a decaying parameter is important as some park users may be just
a fraction beyond ¼ mile, but within the GIS analysis, but this user would be cutout from
the population assessment. It may be more appropriate to use a multi distance buffer as
an approach to evaluate park accessibility across the region (Hass 2009). A multi-
distance buffer analysis could allow for a ranked assessment of users that is more
reflective of the real world user base for a park. With Geographic Information System
software, network analysis can be used in place of buffering as a more accurate means to
measure physical access to a specific park. Road and greenway data serves as route data
to measure the actual ground path from residential areas into the park. Using point of
origins, destination points, and routings, allows ArcGIS Network Analyst to calculate
accurate travel distances between parks and residential areas for use are origin-to-
destination analysis (Esri 2014).
It is also possible to assess the overall distribution of parkland per person as a
method to determine if there are adequate resources to facilitate access (Boone 2009;
Sister Et. al 2009). This may serve as a first step in analyzing the equitable distribution

37

of parkland per person within a given geographic region. Historically this has been used
to evaluate current park and open space inventory and to project future needs on a broad
scale. Having a concise means for measuring access from the parcel to the parks would
allow for quick assessment of availability.
A second challenge to the traditional method of physical assessment is the need to
evaluate the different types of space within the system and to compare these to the overall
distribution. To blanket an area with facilities based on purely a geometric basis will lead
to inequality, as the facilities may not correlate with the target users (Nicholls 2001).
People need to have access to a variety of public spaces to adequately serve their social
and cultural needs (Burgess et al. 1988).
2.5 Summary
In order to understand the parcel-to-park relationships, analysts must be able to
accurately measure whether populations in given residential areas can access a specific
park. Proposed measurement techniques need to be scalable allowing for regional
analysis. How these measurement techniques function in terms of calculating the social
and physical aspects may greatly influence the perceived results of an accessibility
analysis (Hass 2009).
This study is concerned with how the influence of the physical access can limit
the ability of populations to access a given park. While this physical access has a direct
relationship when conducting a detailed park-by-park analysis, GIS-based analysis tools
can allow this physical analysis to be conducted across a regional scale, as demonstrated
below. This would allow for evaluating the overall success of the parks and open space
element of a plan in serving residents across a large area.

38

CHAPTER 3: METHODS AND DATA
This study proposes a method to measure the uniformity of park access by quantifying
parkland available within a given distance to a residential parcel. This cadastral method
will depart from the container and coverage based methods by completing this
measurement for all of the residential parcels within a given community or region. It
would allow analysts to review the uniformity of park access based on the physical
relationship, mitigating the challenges of using more generalized approaches as noted in
the literature review.
Measuring this relationship between parcel and park will be accomplished by
utilizing network analysis procedures to quantify the distance between the physical
access points of the parks and the physical locations of residential properties. Residential
property information is often stored within the cadastral data maintained by
municipalities as a means to track tax information and map infrastructure requirements.
Developing a workflow that is repeatable and scalable with commonly available data will
be key for this technique to be utilized by parks planners and municipal leaders.
The techniques and data used to complete the cadastral-based approach will be
discussed in this chapter. The discussion is organized into the following sections: 1)
Method framework and variable definitions, 2) Network Analyst overview 3) Workflow
development and obstacles, and 4) Methods for presenting results. The first subsection
will discuss the overall framework, describe the key variables needed, and introduce the
formulas necessary to calculating the desired outputs. The workflow development and
data overview sub-section will present the sequence of work necessary to demonstrate the

39

measurement technique including the use of model builder and the manipulation of the
input data necessary to incorporate it into the workflow. The final sub-section will
describe how the workflow outputs were compiled into the final analysis maps and
measurements.
While the use of Esri model builder will be necessary to assist in developing
automated sequences for executing complex and repetitive workflows, the focus of this
study is on testing the practicality of the proposed measurement technique, rather than
developing an application per se. How the various model steps operate with the data will
be described in terms of inputs, parameters, desired outputs, and linked to summary
formulas. Long term, the development of a comprehensive tool that automates the entire
workflow would expand the potential for deployment of the technique across varied
scales such as statewide or citywide analysis; the merits of this are discussed in the
conclusions chapter.
3.1 Method Framework and Variables
The proposed measurement method generates a cadastral-based assessment of the
uniformity of access for parks within Wake County, NC. Successfully developing and
executing the workflow required completing following stages: 1) Extract the residential
areas from the parcel data; 2) Map the locations of park access point; 3) Assign
population to the parcels using census data; 4) Quantify the amount of parkland available
to each parcel per person; and 5) Output the information for review and assessment.
Figure 4 illustrates the proposed workflow, identifying the two main processing tasks in

40

parallel columns. The left column illustrates the work needed to process the park data.
The right column illustrates the processes for the parcel preparation.
The ArcGIS tool Network Analyst was used as the primary tool in evaluating the final
park access. Using Network Analyst requires several steps to be completed that
will be discussed in the following section. Network Analyst was selected because of its

41

ability to work with the existing data within ArcGIS, a common industry tool for
conducting spatial analysis such as park accessibility.

42

Figure 4 Workflow Diagram Illustrating the Cadastral Based Uniformity
Assessment

43

3.1.1 Output Products
The desired outputs of the workflow included both feature data sets and data tables. The
tables were key outputs that provided the summary information necessary to generate the
final accessibility maps based on the compilation of population data, parkland area, and
quantity of parks at a parcel-by-parcel level based on the ¼ mile walking distance. The
cadastral-based apportioning of parkland by parcel allows planners to evaluate park
system accessibility based on the area of parkland available per parcel within the
specified search distance. For this analysis, three primary maps were selected for
rendering: population density, quantity of parks per person and acreage of parks per
person.
Generating these maps required the output data table to be combined with the
parcel feature information through a join process. Joins are a technique within ArcGIS
that unite the data from one table to another based on common attributes. While the three
primary maps were generated from the processed data, a fourth map was also rendered
that utilized the raw data accumulated from the Network Analyst solution of the origin-
to-destination matrix table (ODMT). To construct the ODMT the parcel data needed to
be combined into a summary table that calculated the distances to the parks based on the
original Parcel Identification Numbers (PIN). This summary table tabulated each park
within ¼ mile based on the PIN, allowing for a review of the closest park and total
number of parks to each individual parcel. Summarizing this output table based on the
PIN then created a final output table that totaled the number of parks per parcel and a
total sum of acreage of park per parcel. The latter of these values serving as the final

44

variable to conduct a calculation of acreage available per person based on the parcel
populations.
3.1.2 Data Sources
Utilizing parcel data within the workflow is important in identifying the locations,
walking paths, and distances for residents to reach parks in comparison to the generalized
areal units of census blocks and tracts. While the parcel data in itself may be limited in
the level of detail, the modeling approach accommodates a method for relating the census
data to the parcel data to allow for joining additional data points such as demographics,
income information, and population statistics depending on the type of follow-up analysis
desired. The steps necessary for completing the dasymetric mapping of the population to
residential areas will require a specific subroutine within the model. The following
sections discuss this. Assigning the population per parcel will allow for assessment of
parkland at a per person level if the model is successful.
A key aspect of this study involves departing from a Euclidean geometry based
approach and focusing on the use of physical access paths. Euclidean geometry, as
discussed in chapter two, forms the basis of containers generated from fixed distance
buffers used to represent the service area of a park based on the study distance selected.
The fallacy with such an approach when working within the built environment is
consideration of how travel routes follow the built road network instead of following a
simple point-to-point straight line calculation that may overlook issues such as
geographic boundaries and physical influences such as hydrology and topography. To
measure physical access to parks, it is important to employ network analysis techniques
utilizing physical walking paths that reflect the ground condition in terms of how people

45

will access the park. For this study, the road network will serve as a surrogate for the
pedestrian network to test the model routine.
The core functionality of ArcGIS accommodates the workflow needed to assign
population based on parcel location. The primary tools utilized within ArcGIS are
“select-by-attribute,” “select-by-location,” “join,” and “summarize.” The select-by-
attribute tool selects the features, or objects, based on specific values. For this study, the
target values identify residential parcels with one or more housing units. The select-by-
location tool allows for selection of features that are within a specified straight-line
distance buffer from the source object. The select by location tool also selects target
features that intersect a source feature. The summarize tool generates output tables that
totals, averages, and calculates the standard deviation of the individual features, grouping
them by common features. This study uses parcel identification numbers and unique
object identifiers to facilitate the generation of the summary tables. Beyond theses built-
in capabilities of the base ArcGIS implementation, a network analysis tool is needed in
order to study the travel distances between the facilities and residential properties,
ArcGIS Network Analyst is used to measure route information. The necessary data for
completing this analysis is sourced from Wake County GIS. Table 1 provides an
illustrative link between the source data acquired from Wake County GIS Department
and USDA and the target variables.

46

Table 1 Source Data and Output Matrix
Feature Class Source Variable-fulfilled
WakeProperty Wake County GIS (WakeA
2014)
Household Units
CensusBlockGroupData Wake County GIS (WakeB
2014)
Total Population
WakePublicOpenspace Wake County GIS (WakeC
2014)
Parks
WakeStreets Wake County GIS (WakeD
2014)
Travel Routes
WakeMjrRoads Wake County GIS
(WakeE 2014)
Route Barriers
Aerial Photo USDA 2012 Visual locating of access
points.

3.2 Key Variables and Formulas
The development of the cadastral-based access measurement depends on calculating three
key variables. These variables are total population per parcel, total parkland per parcel,
and total parkland per person. The following subsections provide an overview of how
these variables are calculated. These formulas serve as a foundation for the functions
within model builder.
3.2.1 Total Population per Parcel
Total population per parcel was defined by combining the US Census Data Census Block
Group Data set, CensusBlockGroupData, and the cadastral data maintained by Wake
County GIS Department. The first step prior to calculating the population information
was linking the census data to the parcels using a spatial join. A spatial join allows for all
of cadastral data intersecting the census unit to be assigned that unit’s unique
identification number. This allowed for grouping of the parcels within the cadastral layer
and totaling the household units within the individual census units. The final output is a

47

table of household units per each census unit and the number of parcels within that
census unit.
Calculating total population per parcel requires three variables to be resolved:
total population per block, total household units per parcel, and total population per
parcel. Equation 1 shows a view of how the final calculation is resolved to create a total
population per parcel. Understanding this relationship between the total populations of
the block to the covered parcels allows for an accurate aggregation of population data to
the parcels based on the household units. This approach builds upon on the dasymetric
processes proposed by both Holt (2004) and Maantay (2007), but moves from areal
aggregation to household units. This study also crosschecked the individual block totals
in aggregation to verify that the final total matched the 2010 Census estimate for the
overall population of Wake County.
Equation 1 Population per Household
𝑃 ℎ =
Cp
Th
with variables as follows:
Ph= Population per household
Cp = total population per census unit
Th = total households per census block

To derive the population per household rate, the data needed to be combined
allowing for all of the households within the census unit to be totaled together. Figure 5
shows the model developed to calculate the households per block. Once totaled, the
number of households was divided into the total population of the census unit to derive
the ratio for People per Household for all of the parcels within that census unit. This

48

allows the calculation to scale to the level of detail available regarding population data.
Once completed, the population per household ratio can be applied back to each parcel,
as shown in Equation 2 to calculate the total population per parcel.
Equation 2 Calculation for Total Population per Parcel
𝑃𝑝 = 𝐻𝑡 ∗ 𝑃 ℎ Population variables as follows:
Ht= total household units per parcel
Pp = Population per parcel
Ph = population per household

Figure 5 Model Showing Household Unit by Census Unit

49

3.2.2 Total Parkland per Parcel
Total parkland per parcel is defined by the total parkland area, in acres, within the
search distance of the specific parcel. Parkland area is contained within the Wake
County Parks inventory file and is based off deed acreage. Parkland per parcel is
calculated by totaling the acreage of all of the parks within the search area. Using
Network Analyst, this was a two-stage process to generate the table. The first step was to
running the ODMT to extract a solution table that had the OriginID, DestinationID, and
Length as attributes for all parks within ¼ mile of each parcel. The OriginID represented
the parcel, and the DestinationID represented park access points.
Equation 3 Total Parkland Per Parcel
𝑇𝑝𝑙 = 𝑃 (𝑛 ) + 𝑃 (𝑛 + 1) … 𝑤𝑖𝑡 ℎ 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 :
P(n) = park within search distance of Parcel(i) and is
repeated for all parks found within the search distance for
each parcel.
Tpl=Total parkland
3.2.2 Total Parkland per Person
Total parkland per person will compare the total population per parcel to the total
parkland per parcel providing the analyst with a ratio of the available parkland per person
for each parcel. This will allow a final gap analysis to be generated. While this step can
also be handled utilizing symbology settings within choropleth mapping, it is also
important to extract as an output table for further calculations and for those park planners
not able to work within GIS.

50

Equation 4 Parkland per Person
𝑷𝑳𝒑 =
𝑻𝒑𝒍 𝑷𝒕
𝒘𝒊𝒕𝒉 𝒗𝒂𝒓𝒊𝒂𝒃𝒍𝒆𝒔 𝒂𝒔 𝒇𝒐𝒍𝒍𝒐𝒘𝒔 :
PLp= Total Parkland per person
Tpl = Total parkland per parcel
Pt= Population per parcel
3.3 Application of Network Analysis
There are several options available for conducting network analysis research. Esri
offers the Network Analyst Extension within ArcGIS. This extension provides planners a
means to evaluate multiple questions based on route data and facility information, such as
the quickest route to a location, which locations are within a specific distance of an
origin, and what service areas are available to a specific business (Esri 2014). While
originally developed to conduct retail market analysis for both the logistics and facilities
development aspects, the same criteria can be applied to evaluate park accessibility.
Network Analyst offers two key tools that were utilized in this study: the
“Origin-Destination Matrix Tool” and the “Service Area Calculation Tool.” The Service
Area Calculation Tool generates a polygon based on the specified distance that is
measured from an origin point outward along the available routes at a specified travel
distance. This measurement can be conducted with or without vertical elevation data,
allowing it to function at varying levels based on the quality of the existing data. The
resulting polygon is created by connecting temporary vertices that are generated at the
intersection of the search distance on the available route. The detail, or level of
refinement to the output polygon is controlled within the tool parameters and is useful in

51

modifying how the output shape reflects the surface condition (Esri 2014). For this
study, the refinement was set to 150 feet, allowing for a polygon output that accounted
for the varying widths of right of way and variability in lot sizes. Running this tool for
the entire county created a series of polygons that were used as a preliminary test to serve
as a benchmark evaluating the outputs from the Origin-to-Destination Matrix Tool. The
service area tool was utilized as part of the coverage-based analysis utilized in the study
conducted by Nicholls (2001).
In contrast to the service area tool, the Origin-to-Destination Matrix Tool
(ODMT) generates a set of output lines and a data table that measures the route distance
between a set of origin points and a set of destination points. This tool is primarily used
in logistics planning when determining the influence of how destination locations can be
manipulated to determine where new facilities would best serve the business. This study
utilized the ODMT to map the distance from residential parcels to the parks based on the
location of the park access points. The ODMT offered the ability to link specific parcels
to specific parks.
This approach offered the opportunity to conduct an exhaustive system-wide
analysis of the resources available to each parcel depending on a specified search
distance. This also allowed the study to deviate from the container approach, as it would
allow for a more detailed analysis of real world access from parcel to park as well as the
potential to assess open space at a parcel or even person level.
Network Analyst requires three basic components to run the different routing
solutions. These components are facility locations or origins, travel routes, and
destinations. These different feature classes were loaded into a geodatabase. More

52

details on the development of the geodatabase, including the isolation of park and
residential parcel shape files, are provided in Section 3.4 below.
Once within the database, the files were associated within a network relationship.
The network relationship then allowed the components to be loaded within a Network
Analyst layer for processing (Esri 2014). Prior to loading the data, the facility and
destination locations were converted to point information. The route data needed to be
converted into a network using the integrated data import wizard found within the
Network Analyst Extension (Esri 2014).
Facility locations for both parks and parcels were stored as point information
within the network feature class. Developing appropriate methods for both the parcel
point location and the park information to define these access point locations will
influence the outcomes of both the Service Area modeling and the ODMT. Details on
how park access points were determined are provided in section 3.4.3 below. For
example, single family homes are located within the center of the parcels based on
development guidelines as established in the Wake County Unified Development
Ordinance. Multi-family and Condo developments follow a similar methodology in order
to accommodate required public service and parking space.
3.3.1 Details of Network Analyst
The route information needs to be compatible with the requirements of ArcGIS
Network Analyst. These requirements are focused on how the data is input into the
feature set. Consider how a road network may be digitized into the data set. The road
has a start and stop point. These two points will coincide with intersections. Within
network analysis, these intersections are nodes. It is imperative that the segments for

53

roads are continuous between nodes, without breaks in the lines. It is also important to
ensure that there are no redundant or incorrectly drawn line segments within the route
data (Esri 2014). These elements influence completion of the ArcGIS Network Analyst
tools. The route input wizard will assist the user in checking the data for potential
conflicts in the data for review and validation.
During the execution of the ODMT, there were inconsistencies in completing the
solutions. There were two conditions observed that affected the completion: the number
of possible solutions and inconsistencies in the route data. Repairing the route data to
function within the analysis required using the rebuild option within Network Analyst.
This rebuild was accomplished by making a complete copy of the route data and
removing duplicate and unnecessary vertices and route lines. Crosschecking of this data
was then done by reviewing the provided output file that captured the removed vertices
and lines against the original data set.
3.4 Data Selection
Data used within this study included US Census Bureau Census Block Data, Wake
County parcel data (WakeA 2014), Wake County parks data, Wake County streets data
(WakeB 2014), Wake County major road data (WakeC 2014), and aerial photography
from the US Department of Agriculture Natural Resource Conservation Geospatial
Gateway (USDA 2012).
US Census Bureau Census Block Group data was utilized because of its fine level
of detail and correlation with parcel data, allowing locating neighborhood scaled
phenomena regarding population density and social makeup (Tarrant et al. 1999). The
average block size contains information for between 250 and 500 housing units.

54

3.4.1 Parcel Points
The Parcel data was first culled to locate the residential information. This was
accomplished during the execution of a model that selected all of the residential property
based on the use codes for the parcel. This is shown in which shows the process for
selecting the residential data, completing a spatial join to assign the census data to the
parcels, and generate an output feature class that has parcel number and census id
number. Once this tool was executed, the centroids for each parcel were generated to
create a point file representation of the parcels for loading within the network data set to
use in the ODMT as origin information.
The cadastral information, stored as feature class WakeCountyProperty contains
all of the attributes maintained by the Wake County Tax Administrator’s Office in
addition to the desired criteria for identifying parcel size, housing units, PIN, and current
use. Using the ArcGIS tool select-by-attribute, all of the residential parcels can be
selected from the overall parcel database based on the use codes. Once these parcels are
selected, they were exported into a new feature dataset within the file geodatabase and
the stored attributes to those necessary for park system analysis. The desired attributes
include PIN, Owner Name, Owner Address, Calc_Area, Value, APA_USE, APA_Type,
and TotalUnits. To separate the selected residential parcels from the overall parcel data
set, a new feature class was made from the selection and named “ResidentialParcels”
within the study geodatabase. While this step may not be fully necessary on more
sophisticated computer hardware, this isolation was conducted to improve the processing

55

speed of the workflow by reducing the overall dataset size for use with the existing
computer hardware.
Figure 6 Model for Isolating Residential Parcels and Joining Census Data

The composition of the Wake County development pattern is predominantly
suburban single-family homes. Figure 7 shows a view of Wake County focused on the
area south of Raleigh, NC demonstrating the output of the Points-from-Feature tool
within ArcGIS. These points were used as origins in the ODMT. While it would be even
more accurate to locate each origin point at the road frontage, the use of centroids
approximates where the location of the residential structure is within the parcel. The
smaller the lot, the less influence there is in using the centroid approach.

56

Figure 7 Residential Development Pattern of Wake County, NC

3.4.2 Park Data Preparation
The Wake County GIS department maintains a countywide parks and open space
database and 268 of these features are distinct parks. The parks were isolated from the
other types of open space based on the “park type” attribute contained within the dataset.
Parks were focused on because of their clear assignment, future work will need to review
the classifications of the other types of open space within the data set to add those
features to the park inventory that serve as recreation facilities. Using the resulting
selection set, facility location points were created for the park features to serve as
destinations within the network analysis.

57

Creating the park points was a major decision path in the analysis as not all parks
have street addressing, preventing a simple automated approach for assigning park access
points. This study tested several different methods for automatically creating access
point locations including the use of geocoding to assign points based on the known street
addresses and hand locating the unassigned parks, using intersections surrounding the
parks to generate access points based on proximity of the roads, and assigning points
where roads physically intersected the parks.
City parks will have access potentially on all four sides while suburban parks may
have access only on one side based on both hard and soft edges. While this will be more
visually apparent when running a service area solution through the polygon output, the
origin-to-distance calculation may not reveal incorrect routings. Using the centroid of the
park is not desirable, as it will negate the effectiveness of this study outlining the issues
of physical access through a network-based analysis. Figure shows the workflow
exported from model builder to assign address based access points to a park Access layer.
Figure 8 Model for Assigning Addresses to Non-addressed Parks
3.4.3 Resolving Park Access Points

58

A systematic approach for determining park facility pedestrian access points proved
challenging for this study. Within Wake County, there are several park types typical of a
modern open space system. Like other large counties, Wake County has tight clusters of
urban areas surrounded by large swaths of suburban development. The variation of the
parks includes size, type, and class. Size and context were the two main aspects that
challenged development of a systematic approach.
Size was a major hurdle for using the centroid as a potential destination point.
Using the centroid of the property to place a location point in the middle of a park larger
than 2-3 acres, such as Fred G. Bond Park in Cary, a 3,000-acre park shown in Figure 9
below can lead to false measurement when generating routing from this location point
outward into the community. A park this size has an entry drive that exceeded ¼ mile
just to enter to the park. While this was an extreme example, it required additional
considerations in how the points where located. Beyond the area of cover, the
configuration of the park also had to be considered when establishing access points.
Several linear parks were found within the county; these parks are long and narrow with
diverse access points.
The contextual location of the park was also of concern when locating the access
points. Wake County has a variety of development patterns ranging from rural residential
to a dense urban form. In the urbanized areas within Raleigh, NC, the parks tended to
have roads surrounding the perimeter of the park, often with three to four sides offering
multiple access point opportunities. In contrast, parks that were located out within the
county proper were typically located within a cluster of residential properties with limited

59

access points typical of the cul-de-sac development suburban style. Neither configuration
offered a simple formula to generate appropriate access points.
Figure 9 shows Fred G. Bond Park located in Cary, NC. This park is a large
metro park within the Town of Cary. The access points for this park are located along the
north edges of the park, while the surrounding residential areas are located to the south,
east, and west. Residential parcels that block access to the park from the south, east, and
west sides line the perimeter of the park.

60

Figure 9 Map of Access Points at Bond Park

61

3.4.4 Systematic Approach to Locating Access Points
To assign the access points of a large park like Bond Park shown in Figure 9 , the
access points need to be directly located at the physical entry points in order to capture
the potential users of the park. Erroneous points needed to be prevented by avoiding
false point creation around the remaining perimeter of the park polygon. With the
residential parcels directly adjacent to the park, access along the south, east and west
sides would not be possible without traipsing through private backyards. To account for
this “soft edge” required checking the outputs to verify no false points were created. An
approach like this was required for all suburban, cul-de-sac style parks. The main
entrance point was established and soft edges were eliminated as viable access.
While large parks posed, one challenge when mapping the physical access, so do
linear parks, a type of park that is common in Wake County, NC. An example of this is
shown in Figure 10, demonstrating the pair of linear parks, Forest Park and Wells Park,
within Raleigh, NC. These parks are linear and help demonstrate a second issue when
trying to develop an automated approach to assigning the access points using park
centroid. If this study assigned the park access as the centroid for forest park, there
would be over 300 linear feet from the center to the tip of the park. In a condition like
this, 300 feet represents over a third of the walking distance of a ¼ mile distance.
To combat the problem, access points were defined at the main intersections
surrounding the park, this allowed for a more accurate interpretation of the access
locations when running the solutions through Network Analyst. If this park were located
in a suburban area, limiting access to one or two sides, the number of access points would

62

be reduced to reflect only those points that the public can access, not edges bordered by
private property.

Figure 10 Defining Access for Linear Parks

63

Three steps were utilized to assign park access points while accounting for the
difference in the park sizes and configurations. The first step was to identify the frontage
roads for each park; this was accomplished using the “select by location” tool within
ArcMap to capture all of the streets within 100 feet of the park edges. The frontage roads
were also exported to a new feature class for reference. The “intersect” tool was then
utilized to capture the places where the frontage roads cross the park parcel, the
“output_to_point” option was utilized to generate node points at all of the intersections.
Through exploration, it was possible to expand the search radius of the intersect tool to
capture the road nodes closest to the property corners at edge.
Once the execution of the intersect tool was complete, a second series of points
was added to the newly created access point feature class by using the geocoded street
addresses acquired through the parcel information. The address information proved to be
inconsistent in terms of completeness and required a spatial join with the underlying
parcel data to provide a complete address list. The address points were added to
supplement the initial points as they provided locations associated with the street address
of the parcel, in some cases the midpoint of the street frontage of the large parks, and in
most instances the primary entry points. This step was also necessary, as in several
instances a quality control review of the points was completed, the geocoded point
became the only viable access point in several instances. The access points were stored
with attributes for ParkID to capture the Park ObjectID from the WakeCountyOpenSpace
feature class. This allowed the data to be joined back to the overall database to acquire
acreage information and park name information. Both stages of point generation

64

developed a large dataset depicting over 1,508 potential access points for the 308 park
sites.
Visual inspection of the resulting output was conducted to determine the
effectiveness of the assignment technique. While it appeared effective, several parks
were found to have incorrectly assigned access points. An example of this occurred at
ParkID 92, where the intersection tool captured points at the corners of the park that were
not accessible by car or pedestrian. To accommodate for this, parks were inspected using
the aerial imagery to evaluate the location of the automated points in comparison to the
physical entry points. A new feature class was created titled “Access.” Within this
feature class, the physical access points were added for the parks based on three rules:
vehicular entry points, sidewalk access along the perimeters, and greenway access.
Greenways are pedestrian trails that are located off-road. These trails are not currently
digitized in the same manner of the road networks, lacking the vertices and directional
information found in the base road data. In addition, the base greenway data also lacks
information regarding the entry points to the greenways and locations where the
greenways can access the parks. Locations that were identified during the access point
analysis that appeared to have greenway access were stored within the data set as they
could be incorporated into future work assessing the influence the greenways could have
in the gap analysis and improving access to parks. The greenway features and the
greenway access points were not included in the ODMT, removing them from the results.
3.4.5 Preparing the Census Data
The US Census Block Group Data represents the smallest scale data available for
analyzing population data. This level of detail is desirable when conducting a physical

65

access analysis as it should most closely coincide with the residential areas on the
ground. In order to use the data within the model, several brief processes were
completed. These included calculations for total population per block, assignment of the
geoID to the underlying parcels, and the calculation of people per parcel. The census
block data were assigned to the parcels using a spatial join. The join functioned as a
means to transfer the geoID number of each census block group to each parcel within that
census block. The scale of the block group prevented conflicts where parcels were
located in two or more census blocks. This unique geoID served as the link between the
source population data and the final output features allowing for a consistent link back to
the source data.
Figure 11 shows the model developed to calculate the total population of the
study area to use as a means for cross checking the outputs of the population assignments
to the parcels. Executing the model returns a total population of 900,993 people using all
of the Census Block Data within Wake County. This model provides a means to accept
future data formats that may have multiple instances of the same Block ID. It also
simplified the output table to object ID, GEOID, and Sum_total_population representing
the total population of each block. The Frequency attribute column provides a review of
the features determining if any blocks occur twice.

Figure 11 Totaling Population by Census Unit

66

3.5 Dasymetric Mapping of Population Data
Building on Equations 1 and 2, a series of models were created automating the output
population per parcel calculations. The population data were sourced from the US
Census Data. Within the model, the desired outcome is to have the total population per
census block distributed to the total residential parcels within the specific census block.
This distributes the population to the actual residential locations providing for more
spatially accurate analysis when computing a Network Analyst based origin-to-
destination. This was handled by assigning the census blocks to the residential parcels
through a spatial join. The spatial join integrated the geoID of the Census Block within
the parcels, allowing a unique identifier to tie the block to the contained parcels.
Statistics from the join that needed to complete the final preparation calculations
included the join count, total population per block, and the total parcels per block. Using
the “summary statistic” tool allowed for these items to be calculated and added within a
revised attribute table for the residential parcel data.
The majority of housing within Wake County, NC is single-family homes.
However, there are also concentrations of multifamily homes located in clusters
throughout the county. As part of the data management, the Wake County GIS
Department, in conjunction with the Wake County Tax Office, monitors the number of
housing units on a parcel level. This is a key metric for the Tax Collector to assign tax
values and assess fees to landowners within the county. Using this information, it is
possible to account for the differences in housing type when applying population
calculations. Within the parcel data, the use of the housing units as the means for
calculating the population reduces the risk of the MAUP as opposed to using parcel area.

67

This is because the population is calculated by using the direct ratio of people per unit as
calculated per census unit to the known total of household units. This mitigates the risk
as the application of average people per household of that specific block to the actual
housing units of the parcel, as stored within the Wake County cadastral information,
allows this study to account for both single family and multifamily housing as opposed to
those studies that aggregate population based on land area derived density calculations.
3.5.1 Final Output Table
The final output table from the population mapping contained the core information for
population per parcel, geoID for the census blocks, and the total number of households
per blocks, and the PIN. These attributes are what were used at the completion of the
ODMT to create a final summary table. Figure 12 shows an illustration of the table of the
residential parcels with the assigned population per parcel calculated.
Figure 12 Illustration of Summary Data for Residential Parcels with Population
Calculated
3.6 The Origin to Destination Matrix Tool Solution
Once the data preparation was completed in the subsections above, it was loaded into the
ODMT working layer. Origins were loaded from the residential parcel data, destinations
were loaded from the Access layer created for the parks, and the Routes were added from

68

the revised route information. Barriers to the routes were added from the WakeMajorRds
dataset.
This created a total network dataset that had 276,860 residential parcel origins,
672 distinct destination points for the 268 unique park features, and over 8,668 barrier
features for the routes. The parameters selected for the ODTM were left bi-directional as
pedestrians can use the sidewalks in any direction as opposed to limitations in vehicular
circulation. The search distance for identifying nodes that could be used in the ODMT
solution were limited to a 2 mile search pattern, or 10,560 linear feet. This larger search
area was utilized in order to make sure that the ¼ mile distances were not lost due to
cutoffs in the network solution.
To crosscheck the output of the ODTM an equation was developed to verify that
the solution outputs were within the predicted outcomes where the total should equal one
mapped route for each possible destination from each origin.
Equation 5 ODMT Output Check

𝑂𝐷𝑇𝑀 𝑆𝑜𝑙𝑢𝑡𝑖𝑜𝑛𝑠 = 𝑂𝑟𝑖𝑔𝑖𝑛 ∗ 𝐷𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛
With variables:
Origin = Total origin points input into the ODMT
Destination = Total park access points input into the ODMT,
excluding greenway access points

This total was based on the number of destinations found within the ¼ mile search
distance, calculated using the service area generated polygons. This allowed for an
estimate to be completed of 17,965 origins within ¼ mile of a park. As the service area

69

tool also was grounded in the same route data, it returns a similar number of destinations
when the service area polygon is used to select the residential parcels it overlays.
Once the ODMT completes a solution, there are three outputs stored within a
ODMT layer within the map document. These output classes are a copy of the origins, a
copy of the destinations, and the lines generated to represent the origin to destination
paths. These lines, depicted as straight lines between the points, are used to visually
illustrate which points have been connected to the origin. To complete the final output
table, a series of joins are required, associating the target data information to the now
generated distances.
Step 1 was to join the park Access data back to the destination data. This join, or
uniting of data tables, is accomplished using the ParkID from the Access file and the
Name attribute of the destination features. Once this was completed, the destinations are
now mapped to park name and park acreages.
Step 2 joined the revised destination table to the lines table using the
DestinationID stored within the lines feature class to the ObjectID within the Destination
class. This created a table that now contained the DestinationID, ParkID, Park Acreage,
and Lengths.
Step 3 was to join this revised line table to the Origin features using the OriginID
within the Lines feature and the ObjectID within the Origin features. Once complete, this
final table was exported to a database file table (DBF) and added the map. This DBF
contained a breakdown of attributes for Length (between origin and park access point),
ParkID, ParkArea, and ParcelID. This table was then summarized to create a refined

70

summary table that had the total park acreage and count of parks within ¼ mile to each
origin. This final table was saved as ParksPerParcel.dbf and added to the map.
By joining this final table to the original cadastral data, the final result was an
updated parcel feature class that contained all of the parkland assigned to all of the
parcels as found within the ¼ mile search distance. This allowed for the generation of the
final review maps for Parks per Parcel, Acres of Park per Person, and a Population
Density Map based on the cadastral information. The lines file, once joined to a second
copy of the cadastral data was used to generate a gap analysis map that showed the
closest park within 2 miles for all of the residential parcels within Wake County, NC.
3.7 Test Hardware and Software Setup
The testing for the workflow was accomplished on a Lenovo Thinkpad W530 portable
workstation. This system was equipped with a Samsung ev840 mSATA system drive and
a 500 GB Samsung ev840 SATA data drive. The workstation was powered by a single
Intel Core i7-340QM processor running at 2.7 GHz. The processor was supported by 24
GB of RAM to aid in completing the computations. The system was running Windows 7
Professional 64bit with Service Pack 1 installed. The graphics card installed was an
NVIDIA Quadro K1000M mobile graphics processor. This laptop was purchased in
2013.
The initial phase of this study was started within ArcGIS 10.2.2 Advanced
Desktop running ArcMap with the Network Analyst extension and ArcCatalogue. To
improve performance and success rates, ArcGIS 10.2.2 was updated to ArcGIS 10.3
Advanced Desktop prior to executing the final testing of the workflow and production of
the output maps. Computations were completed using the Background Geoprocessing

71

tool for 64-bit systems, also available from Esri as part of the ArcGIS install platform.
All map production and table generation was accomplished within ArcGIS Map and
Catalogue. PyScripter 2.7 was used to check the script outputs from model builder to aid
in finishing the calculation portion of the workflow.
While more advanced computing resources were available at the time of this
study, it was important to test this workflow on a computer that most likely would equal
what the typical park planning staff may have at hand. The laptop, classified as a
workstation, utilizes a standard mobile processor that is common on many desktop
computers and consumer grade laptops.
Microsoft Excel and Microsoft Word 2014 were utilized to generate the report
tables once the primary formatting was completed within ArcGIS. Adobe Photoshop CC
was utilized to combine the individual maps into the exhibits presented within this report,
with no modifications or manipulation of the color ramps stored within the choropleth
maps.

72

CHAPTER 4: RESULTS
The goal of this study is to develop a measurement technique that will best determine park
accessibility based on the real world ground conditions for both populations and for travel
distance. This measurement technique focuses on population location linked to access points for
the parks utilizing access points that have been mapped based on physical connectivity on the
road network. The connectivity of the road system within the data is directly related to the
ability of the ArcGIS tool Network Analyst to execute potential solutions. The proposed method
depends on the Network Analyst extension to complete its solution in order to generate the
required tables and features necessary for evaluating the accessibility of the park system.
This chapter reviews lessons learned in applying Arc GIS Network Analyst to develop
the cadastral-based accessibility measurement technique. The study results also include
comparisons in levels and locations of park access for the Wake County, NC study area between
the origin-to-destination matrix table (ODMT), a buffer based coverage analysis, and the Service
Area Calculation tool (SAC) provide a means to quantify the improvements from using Network
Analyst in measuring access.
4.1 Lessons Learned in Applying Network Analyst
As mentioned in Chapter 3, attempting to calculate network distances for all parks and parcels in
Wake County, NC using the ODMT tool quickly exhausted the computational resources as the
number of possible origin-to-destination (OD) pairs, according to a calculation similar to
Equation 5 in Chapter 3, were in the millions. ArcGIS did return results, but a check of the
resulting tables, showed them to be incomplete. It is likely that ArcGIS simply stopped
processing when memory capacity was reached, but it is important to note that no error messages
were returned.

73

To better steward the computational resources, it was necessary to first complete the
service area assessment across the county using the Network Analyst Service Area Calculation
(SAC). This tool as noted in the methods generates points based on the distance along available
routes from a given destination and then connects the closest vertices together to create the
polygon. Settings within the SAC allow for the control and refinement of how this polygon is
created. As noted in the methods, the detail level was kept high and the search distance was
limited to 100 ft.
Completing the SAC was able to be accomplished on the test hardware and software. The
overall process for the Service Area Calculation did not require any deviation from the stated
methods. At the completion of the solution, 284 park sites were represented within the SAC by
the 636 access points located through geo-referencing of the street addresses and locating
through aerial photography for each park. The SAC tool was run using the three distances of
1320, 2640, and 5280 feet, with feet being the calculated length provided in the Network Analyst
outputs. The distance cutoffs represent ¼, ½, and 1 mile distances. While the focus of the
analysis is on the ¼ mile walking distance, it was of interest to this study to gauge the
performance of the SA tool and explore how its functionality could be further integrated into the
gap analysis workflow.
Using the three distances, one can compute the number of service area polygons according to
Equation 6.
Equation 6 Total Paths Possible ODMT
𝑇𝑜𝑡𝑎𝑙𝑆𝑒𝑟𝑣𝑖𝑐𝑒𝐴𝑟𝑒𝑎𝑠 = 𝑛𝐷𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛𝑠 ∗ 𝑛𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒

Destinations= Number of Sources
nDistance = the number of distance cutoffs input into the variables

74

Calculating values similar to Equation 5 and Equation 6, although simple, also proved necessary
as a means to conduct quality assessment checks on the OD matrix outputs. As the number of
facilities being assessed increase, so does the complexity of the computations as well as the size
of the map file in terms of file storage. Issues in file size are further discussed in Chapter 5
below.
For Wake County, the use of 3 distance cutoff values and 636 access point’s results in a
total of 1,908 polygons. These polygons are then able to be used to conduct a coverage based
selection of the residential parcels within 1 mile of the park access points based on the road
network. In comparison to calculating links for all parcels and parks in the county, this use of the
SAC greatly reduces the number of potential line calculations and allows for the ODMT routine
to complete when generated in conjunction with the barriers.
4.2 Determining the Parcels Served
The parcels served is the first stage in completing the gap analysis. Through the development of
the methodology, the SAC was utilized as a means to conduct comparisons between the buffer
based coverage analysis and the ODMT outputs. The generation of the SAC also served as a
visual check to assess which sites were potentially within the specified distance of the park
facility by creating the polygons at the specified search distances of ¼ mile, ½ mile, and 1 mile.
The limitation of the SAC is that it is a singularly oriented process in which the service area is
defined outward from the park access points.
The SAC also allows for the assignment of distances to the lines traveling from the
access point for the specified distance. These travel lines are an overlay of the roads following
the search distance criteria and provide for a visual check of the road network, confirming its
functionality prior to being used within the ODMT. Adding these lines to the network analysis

75

layer provides the analyst the ability to display the routes from the parks outward using a color
ramp. Using the roads after verifying connectivity within the SAC also improved the success rate
of the ODMT calculations as breaks within the road network were quickly identified. A second
advantage for generating the service area is the SAC tool updates the spatial information within
the network for all of the park points allowing for computational efficiency and a corresponding
reduction in the processing time within the ODMT. Figure 13 shows the output of the Service
Area tool for all of Wake County, NC. The processing time for the overall county was less than 5
minutes and proved to be replicable using all of the access points for the parks.
One of the key criteria established in the methodology is a park’s ability to be walkable. With
walkability estimated to be manageable at ¼ mile, this will be the primary metric for evaluating
the service coverage of the park. Using the service area as a visual assessment, it is readily
apparent that there appear to be large gaps in the coverage between the areas of park coverage.

76

Figure 13 Service Area Map Based on Service Area Calculation

77

4.3 Output from the OD Matrix Tool
Completing the ArcGIS Origin Destination Matrix Table (ODMT) for a countywide analysis
requires working within the limitations of the computer memory available and file size
limitations imposed by ArcGIS. This study found that a maximum search distance of 2 miles
processed before the ODMT stopped returning complete solutions. Using this limit in
conjunction with the barriers allowed the ODMT to execute successfully. A secondary
improvement on the ODMT was using the origins and routes utilized in the ArcGIS Service Area
Calculation Tool as they were previously located and improved the computing speed in resolving
the ODMT outputs using the origins and routes mapped in the SAC.
The ODMT returned a solution that measured connection lines for all of the parcels
within 2 mile access of each park. Using the select by attribute feature, those origins located
within ¼ mile of the park were able to be isolated to just over 31,000 origin-to-destination
outputs. Once summarized, 17,795 unique parcels had park access within ¼ mile. The results
were limited to those points that were located along a continuous route to the park. Routes that
were blocked by the barrier led to exclusion of those origins to the destination park. Figure 14
provides an illustration showing the ODMT output running within ArcGIS, with the mapped
lines representing the solved OD pairs and the redlines showing the barriers. The table on the

78

right of the map shows the typical output of the lines once joined with the Origins attributes
linking ParkIDs back to the line data for analysis.
The raw table allowed for several quick assessments of the park system. The first and
most simple was to summarize the data based on the ParkID. This allows for a chart and graph to
be generated to depict the number of parcels served by a specific park, average distance between
the park and the parcels, and the number of parks that serve a particular parcel. An example of
the output is shown in Figure 15.

Figure 14 Raw Output from ODMT within ArcGIS

79

Figure 15 Output Table from the OCDM Calculation
The ODMT chart is joinable to the original parcel data. A comparison of the number of
parks by parcel and the amount of open space by person was accomplished through the joining of
the data based on the common attribute of the originID and the objectID.
In Figure 16 Population Density per Parcel, hard edges defined by transportation
corridors, and soft edges created by areas of non-residential development break up, or segment,
the residential areas into isolated clusters. It is also of note how areas of the highest population
concentrations are located at main thorough intersections around the edges of the urban core.
Generalized density mapping approaches overlook this type of density distribution when using
larger areal units.

80

Figure 16 Population Density per Parcel, Overall County

81

Figure 17 Population Density per Parcel, Focus View

82

Figure 18 and Figure 19 illustrates the number of parks by parcel within 1/4 mile
showing that there are areas in the urban core have more parks available to them in comparison
to the suburban parts of the county. Refer to Figure 18 to Figure 16 to compare the difference in
number of parks per parcel to residential density. Assessing the difference in values between the
residential clusters and the clusters of parcels with the greatest quantity of parks will reveal
issues with uniformity of access. The network barriers significantly influence this output.
Adding proposed access improvements to the network dataset, such as bridges over the major
roads, updates the ODMT solution; this allows for evaluating one or more modifications to the
park system to determine the potential benefits through additional parcel connectivity. Similarly,
adding parks to the areas of lower coverage would improve the overall availability of park
facilities. The ODMT allows for on the fly testing of such scenarios by allowing for new
destinations or network paths to be added and updating of the output table.

83

Figure 18 Parks per Parcel, Countywide

84

Using the ODMT output information, a third map was created depicting the area parkland
available per available per person for each parcel. This map shows the variation in the amount of
Figure 19 Park Count within ¼-mile, Enlarged View

85

parkland that occurs across the county. This allows for a clear depiction of where there are
shortages in total area available to each parcel, an accepted standard in planning park spaces.
Figure 20 and Figure 21 depict the acreage of parkland available per person as derived from the
cadastral data and the ODMT outputs. Figure 20 is a view of the output at a countywide scale,
Figure 21 is a view an enlarged view focused on the parcels near Raleigh, NC.

86

Figure 20 Acre of Parks within ¼ mile of Parcel, Countywide

87

Figure 21 Acre of Parks within ¼ mile of Parcel, Enlarged View

88

4.4 Comparison of Accessibility Measurements
Using the ODMT to generate a countywide solution processed 687,651 solutions for the original
276,860 origins and potential pool of 636 access points; using a 2 mile search distance. The
access points, as noted in the SAC calculation represent 284 park sites. The max possible results
of a completed ODMT solution would reach 78,628,240 Origin-to-Destination pairs when using
Equation 1. However, running the OD solution using the barriers (i.e. the major roads) this
number drastically reduces. To estimate the potential output, the SAC polygon intersected
14,095 of the 27,680 parcels when using the break distance of 1,320 ft. or 1/4 mile. The SAC
number represents the intersection of one parcel to one park service area, for a one-to-one
relationship. While the SAC is useful in assessing service gaps, the ODMT tool will return a
one-to-many relationship between each parcel to all parks (access points) within the specified
search distance. In instances of multiple access points for a single park, the ODMT selects the
closest access to the parcel (Esri 2014). In contrast to the SAC estimate, the ODMT returns a
value of 31,963 parcels to park relationships. The discrepancy in these two values shows that
multiple parcels have two or more parks associated within the search distance. This is the
strength of the ODMT tool in providing highly detailed results with limited post processing on
the data. A secondary benefit of the ODMT is the ability to identify the closest park for each
parcel based on the calculated OD pair lengths.
The ArcGIS Summarize tool isolated the unique instances of the parcels, totaling the
amount of park acreage per parcel instance, and counting the number of instances of parks that
were within the target search distance. The summarized parcel count is 17,795 parcels located

89

within ¼ mile of a park access point. The summarize tool also returned the first of the vectors
being tested, gauging how many parcels have access to more than one park. Based on this
study's results the park access for parcels is highly skewed: 6,473 parcels that have access to
more than one park. In contrast, this study shows that 259,065 residential parcels are not within
1/4 mile of a public park based on the available routes.
The overall output of the ODMT provides a high level of detail in terms of understanding
the relationship of parks to parcels. The following set of tables illustrates the comparison
between the ODMT output and previous methods. In Table 2 Review of Residential Parcels
within Wake County, NC, there are 276,860 residential parcels, or roughly, 82.23% of the county
has residential features.

Figure 22 Summary Statistics for Number of Parks per Parcel for Wake County,
NC

90

Table 2 Review of Residential Parcels within Wake County, NC
Description Parcel Count Percent of Total
Total Wake Parcels 336,667 100%
Residential Parcels 276,860 82.23%

Table 3 provides a summary table showing the difference in residential parcels identified within
¼ mile of a park. The Service Area Solution uses the ¼ mile search distance to generate a ¼
mile service area from each of the access points for each park. As explained in the methods
section, the SAC uses road network information to locate vertices ¼ mile from the access point
based on travel path. Parks with multiple access points generate multiple service areas, these
were converted into single service area polygons for each park using the ArcGIS Join tool. Once
complete, all of the parcels within the service area polygons were selected and totaled in the
table below.
The ODMT total was collected from the output data. The interesting point of note is that
there is less than 0.3% difference between the SAC output and the ODMT output. The notable
difference is in comparing a traditional ¼ mile buffer based coverage analysis that shows over
26.27% of the residential parcels are within ¼ mile of a park. This is much higher than the
cadastral-based approaches, either the SAC or ODMT models, which show between 6-7% of the
residential parcels having access. This difference in levels of accuracy quickly illustrates the
value in using a cadastral-based approach to measuring park access.

91

Table 3 Comparison of Selection Methods
Vector Method Count Percent of Total
Residential Parcels 276,860
Residential within ¼
mile of Park
Coverage (SA
Polygon Hybrid)
18,627 6.72%
Residential within ¼
mile of Park
ODMT 17,795 6.42%
Residential within ¼
mile of Park
Coverage – Radial
Buffer, Park Edge
72,735 26.27%

A histogram generated from the ODMT provides a graphical representation of the distribution of
the acreage of parks available per person. Figure 23 shows the mapping of the parcels that are
located within ¼ mile park, ranked by the amount of park acreage available. This type of
analysis shows that there are areas within the park system that have more parkland available to
concentrated areas of residential development compared to the overall distribution.
Figure 23 Histogram Graph of Park Acres per Parcel

92

A final map generated from the line data feature of the ODMT may prove to be the most
compelling reason to utilize the ODMT instead of the SAC based analysis. The ODMT
measures the distance between each origin to each destination, only stopping when it hits the
specified search distance or quantity of returns desired. By storing the distance between each
parcel to each park within a data table, completing a closest facility analysis to evaluate the
trends in travel distance across the park system is possible.
Using the travel distance information stored in the ODMT data table facilitates a
uniformity of park access assessment showing the minimum, maximum, and average travel
distance. Figure 24 is a histogram graph displaying the trend of travel distance using the closest
park per each parcel found within a 2 mile maximum search distance.
Figure 25 and Figure 26 use the ODMT to generate choropleth maps identifying where
there are parcels that are quite close to a 1/4 mile cut off, but do not quite reach pedestrian access
by the 1/4 mile rubric. Figure 25 is a countywide view of the distance between parcel and
Figure 24 Histogram Showing the Variation in Distance to Parks

93

closest park using the full ODMT search of 2 miles. Figure 26 shows an enlarged view of the
parcels near Raleigh, NC. These maps would allow planners to search for physical gaps in the
access using the direct output from the ODMT tool. This is an important distinction from the
SAC because it will allow for quick assessments in terms of how new park sites and additional
access paths (e.g. public rights of way, overpasses, or nature trails) can improve the overall
functionality of the system. To do a similar analysis using the SAC, additional work would be
necessary to regenerate and identify where the gaps area.

94

Figure 25 Maps Showing Park Access Based on Nearest Distance to Parcel

95

Figure 26 Maps Showing Park Access Based on Nearest Distance to Parcel

96

CHAPTER 5: DISCUSSION AND CONCLUSIONS
This study set out to develop a technique for measuring and assessing the physical accessibility
of an existing park system in order to map the accessibility of parks on a parcel by parcel, in this
case Wake County, North Carolina. To resolve the three questions asked in the initial onset, the
study completed a series of trials in order to arrive at the best combination of variables to
generate the most useful output. This study has revealed the potential benefits for using this
measurement technique to evaluate park accessibility. Developing this process has also helped
identify potential issues in automating this process and areas meriting future research.
5.1 Benefits of the Cadastral Based Measurement Technique
While there are numerous aspects to siting a park, the ability of target populations to access a site
is a key metric for evaluating the potential benefit. The cadastral-based method proposed here
uses a series of processes to combine population data, household data, park access points, and
streets into a final output that totals the number of parks and quantity of park space available
within the specified walking distance. The search distance is adjustable depending on the mode
of travel allowing for a comprehensive assessment of the distribution of parks and a quick
assessment of the park coverage to identify the gaps in the system.
Within this study, the ArcGIS Origin-to-Destination Matrix Tool (ODMT) proves
capable of measuring the number of parks available to each parcel based on the route data within
the study area. Adjusting the variables within the ODMT allows the analyst to tailor the output
to focus on search distances based on study criteria. The initial output of the ODMT is a matrix
table and a line feature class that maps each parcel to each park. The use of choropleth mapping
and histogram charts based on the distance field of the travel lines assigned to the parcels
provides a quick visual illustrating where the gaps in coverage exist. More importantly, once

97

generated, the ODMT solution is stored within ArcGIS as part of a geodatabase, or container file
that holds all of the information necessary to rerun the evaluation. This allows quick adjustments
to access points or origin locations, followed with an updated ODMT solution utilizing less
computing resources.
The ability to rerun the ODMT using updated data and new origins or destinations
quickly is a second benefit as it allows the tool to be interactive beyond the initial analyst. Park
planners can modify access point locations, add new locations, and update the underlying parcel
information as new residential areas develop; quickly evaluating how these changes influences
the overall uniformity of access.
The advantage of this output is the user can combine the different components through a
shared attribute, such as the originID, to evaluate the entire data set or specific parcels.
Exploring the data output also shows an advantage over other service area coverage approaches
as the origins can be sorted and summarized by instance, or conversely the travel distance can be
used to further refine selection sets within the study to evaluate trends. One example is
searching for the number of parks that serve more than one parcel within 1/8 of a mile; using
standard query expressions within ArcGIS or even through the database server will allow
analysts explore the nuances in access and site locations.
5.2 Disadvantages of the Cadastral-based Technique
Conducting Origin-to-Distance Matrix Tool (ODMT) solutions creates an output table that
exponentially increases in relationship to the number of origins to destinations. There are
definite limitations in ArcGIS Network Analyst due to computer memory, computer processor,
and data storage and file size limitations. Several tests were conducted countywide utilizing all
of the residential parcels evaluating the performance thresholds. For reference using an ODMT

98

for all of the over 286,000 residential parcels and 308 parks (with 626 access points) can create
upwards of 88 million separate lines when using a simple 1 parcel (origin) to 1 park access point
(destination) calculation. The output data tables for the calculations using just a 1/4 milesearch
distance exceed over 600,000 different returned solutions, expanding this to a regional approach
would exponentially increase the output. Running the ODMT for all of Wake County pushed the
limits of the ArcGIS file size, with the output map files exceeding 1.2 gigabytes; this appeared to
be the limit of ArcGIS.
5.3 Origin to Destination Matrix Tool Compared to the Service Area Calculation
One product of this study is the comparison of the ODMT tool to the Service Area Calculation
(SAC), both of which are within ArcGIS Network Analyst. The output of the SAC selected a
similar number of parcels using the ¼ search distance. In addition to reaching a similar number
of served parcels, the SAC was able to generate the output areas using the same destination and
route sets as the ODMT. While there is a tangible benefit for increased processing speed using
an SAC in place of the ODMT when regarding multiple assessments, there are several key issues
to consider.
The Service Area Calculation tool generates a polygon that is useful within a coverage
type approach. The strongest advantage of the SAC polygon is its foundation based on travel
distance along the road networks, reducing the generalization found within traditional buffers.
Beyond this advantage, incorporating the SAC within coverage-based evaluations, the analyst
must define the search criteria, this variable will modify the output numbers of the selection set.
Setting this variable too high or low will the benefit of the detailed routes is potentially lost.
This risk, while using a more realistic service area as opposed to buffers, may be acceptable
within certain studies depending on the allowance for tolerances.

99

The output from this type of coverage analysis also results in a one-to-one selection
between park and parcel. Modifying this output to allow for tallying total parks available to each
parcel and calculating the acreage requires a large amount of interaction. The ODMT selects all
features within the specified travel distance along the navigable routes, creating a one-to-many
output that reduces the amount of manipulation needed on behalf of the analyst to derive the
same output from the SAC coverage based analysis. The ODMT output assigns the calculated
length, parkID, and occurrence information to each parcel, something the SAC also cannot do
without additional manipulation by the analyst.
While computationally more taxing, the ODMT proves to require less interaction to
generate the most comprehensive set of output data. A key component in any assessment is
choosing a tool that requires minimal input and generates the most useful data for the target
analysis. Reducing the number of steps within a process will potentially reduce the opportunities
for incorporating sources of error; equally important is minimizing the number of variables that a
user can modify in order to return consistent results.
5.4 Future Research
The overall process of using the combination of dasymetric mapping and ArcGIS Network
Analyst as part of a parks planning workflow reveals several areas that warrant additional
research. These additional areas of research would evaluate data acquisition, data standards,
workflow streamlining, and improving the functionality of the process through automation.
Automation through development of a plug and play style geospatial application could make the
cadastral-based measurement approach accessible to city planning departments with varying
levels of GIS expertise.

100

The quality and completeness of the input data can greatly influence how the model
functions. The standards that govern measuring, naming, and acquisition scale directly
influences the quality of the data collected. This study is possible in part due to the level of
detail maintained by Wake County GIS department for cadastral information, road networks, and
the inventory of public and private open space.
For example, within the cadastral data for Wake County, NC, there is an up-to-date
attribute field for tracking the number of housing units within a given parcel. This attribute
allowed for the reassignment of population at a very large scale across the county. This cadastral
data also maintains attributes that track the year built, site function/use, land value, and acreage.
All of these attributes can lead to a more comprehensive analysis on park accessibility such as
limitations in moving closer to a recreation area due to land costs or lack of available housing
units.
Population census data is currently on a 10-year renewal cycle within the United States.
Conducting park service analysis needs accurate population projections to predict how the
density of a given area may change, create pressures on existing parks and guide future park site
selections. The use of an ODMT to assign park distances across the study area creates an output
feature class that may integrate within a predictive population model to assess redeveloping areas
to identify areas with potential future park shortages. Such a model would be plausible if
population modeling uses a constant like housing units found within the cadastral information.
The dasymetric model documented here provides a means for apportioning future census data to
the cadastral data. With the potential for census blocks to change in area as populations increase,
this linkage would require some careful thought.

101

The complexity of integrating the population data and parcel data in the base form of this
work into a completely automated process is also worth additional research. One of the initial
intents of this study was to develop an automated tool, as least as implemented through model
builder. During the model development process, an automated process for generating the final
population tables and the park access points was not completed. The processes for the
dasymetric mapping component is automated through the use of tools developed within ArcGIS
Model Builder including calculating the census block population numbers, assigning census
block ids to parcels, totaling of household units per census unit, and total population for each
parcel.
5.4.1 Adding Social Variables
This workflow has been tested using existing data for population, household units, and park
locations based on a physical road network. These vectors are quantifiable and mapped within
existing data. Advancing this workflow to incorporate the social variables associated with
demographics, public opinion, and current user statistics would allow for further analysis of how
the differing social groups within the study area have access to the parks. Considering this
interaction, there are two directions to evaluate, looking at the parcel level to create a clear
definition of how the different groups can access the park independently, and at the larger scale
using neighborhood and social boundaries within the ODMT calculation to understand the
influence of perceived soft edges in regards to park distribution.
This latter concern could reveal an even larger level of fragmentation within an open
space system as neighborhood demographics can serve as a source of perceptions including
safety as well as social acceptance when crossing these soft edges to utilize a park in a different
neighborhood. Mapping safety by adding in point instances of crime tied to the same census

102

block group data, providing a more tangible set of data points to create a calculable result.
Incorporating social perceptions would require a more comprehensive approach that may include
gathering direct survey data from residents regarding parks they use across the county,
identifying the parks not visited due to social factors separate from measurable travel distance.
These surveys, sampled from each of the different racial and income groups, would provide
insights in mapping potential social barrier lines when incorporating perceived cultural and
emotional barriers as realized barriers within the ODMT.
5.4.2 Automation and Portability
Fully automating the process is a second focus of future research. At the start of this study,
several stages of the data management and dasymetric process were automated through model
builder. Due to limits in knowledge of scripting and advanced data management, further study
and exploration would allow for a complete model to be developed that can function regardless
of geographic location. This revised automated model would allow analysts to input their source
data for parcels, population, park access points, and barriers and execute multiple solutions with
minimal input within the workflow.
Beyond automating the main workflow, developing an automated approach for
determining access points for subject parks would be of tremendous value in improving the
practicality of this approach across multiple regions. For Wake County, while the geocoded
addresses provided a basic level of access points, the field conditions showed that many parks
have multiple access points that required further mapping. Omitting these access points from the
ODMT calculation would create false gaps in the coverage area. Manually inputting these points
for over 300 park sites requires a commitment of time in both the field and through remote

103

analysis comparing different data layers and aerial photography in order to make accurate
demarcations.
The quality of the geodatabase for the road network is also of concern as the scale of the
study area increases. Having a cleanly digitized road network with concise locations for
intersections, sidewalk information, locations of pedestrian crossings and speed limit attributes
would allow for higher accuracy and improve the ability for the ODMT to complete successfully.
In the initial work-through, the calculations for the ODMT within Wake County had inconsistent
completion rates. Improving the completion rates required several changes. In ArcGIS Network
Analyst for example it was necessary to use repair and re-build tools to strip out redundant
vertices and overlapping lines. However, this tool can lead to oversimplification, requiring the
analysis to review the output network to verify its completeness.
The use of roads as barriers is also a crude method when considering the proximity of
parcels to the road and how the ODMT handles parcels that are on the same side of the road
along a barrier as the park. This will require additional research to evaluate methods for
automating the identification, digitization, and categorization of sidewalks along existing travel
routes. Once identified, these sidewalk routes could be incorporated into the network dataset to
create a more complete route network for use in the ODMT calculations.
5.4.3 Workflow Accessibility Evaluation
Integrating ArcGIS Network Analyst into this study limits the initial user group capable of
executing this workflow. Limitations include skill set, cost of hardware and software
(acquisition and maintenance), and the availability of the required base data. Distribution of an
automated tool will mitigate the skills gap for those communities and organizations that that have
access to the appropriate software licenses.

104

Mitigating the software and hardware costs requires further evaluation. Computer power
and equipment costs will continue to improve, however the cost of operating and maintaining the
hardware continues to increase. There are options that are developing within the geographic
information systems industry that may prove viable as alternatives (Lwin et al. 2012). These
opportunities include the advancement of online-based GIS services such as ArcGIS Online.
The Trust for Public Lands maintains a system branded ParkScore that may also support the
proposed workflow.
The availability of Open Source alternatives to ArcGIS and ArcGIS Network Analyst is
increasing, candidates that include Grass and QGIS both support network analysis type
workflows, but require learning the software. Developing an automated script that works within
these platforms would allow communities that have an existing GIS infrastructure to utilize their
data to conduct this type of workflow, while avoiding having to purchase ArcGIS and Network
Analyst.
Improving base data is a more complex challenge when considering the need for detail
regarding routing information and parcel data. In North Carolina, parcel data is comprehensive
at the county level, in order to support the tax collection infrastructure. In other states, this may
not be the case. The use of crowd-sourced data such as that from the OpenStreets project may
improve the route analysis as users can add information regarding sidewalks, overpasses, and
other components to the data set that will greatly improve the quality of the output.
5.4.4 Scaling Workflow
Testing this methodology on a large county within North Carolina that is representative of the
level of complexity expected within the state in terms of parcel count and total parks serves as a
means to evaluate the potential limitations in equipment and software. Wake County, with its

105

mixture of urban and suburban development areas also serves as a viable test case for
experimenting with the road network and necessary level of detail for resolving detailed origin-
to-destination measurements. This study used the walk distance variable as a means to cull the
potential number of Origin to Destination pairs within a single county. Despite the observed
limitations in computer processing and file size, this workflow may be scalable through one of
two methods that will need to be evaluated further in order to address larger more developed
counties, such as Los Angeles, and how to account for multi-county scenarios common in
regional assessments.
Successfully implementing the automation of the proposed approach will be a primary
step in refining this approach to work at the larger scale. In order to handle larger counties, the
automated script (or model) requires modification to evaluate a fixed area within the overall
study area. Once the first area is complete, the focus area would shift to analyze the next set of
origins, processing the larger county in a grid. This technique compartmentalizes the county into
smaller segments that are manageable within the limitations of the hardware and software. Once
the ODMT solutions complete for each of the focus area, a script needs to collect the output
tables into a large unified table via a database management program capable of handling file
sizes larger than ArcGIS, such as Microsoft Access or within the SQL server. Within the
database program, one can summarize the ODMT data and merge the output to the parcel
database. This allows analysts to complete the uniformity analysis for the larger county.
A similar approach will solve the multiple county scenarios. The consistent link is the
use of the automation script to facilitate the larger volume of source data. The challenges that
face this scenario include consolidating the park information, the parcel data, and acquiring a
unified route database. The route database is of significant concern, as the roads need to connect

106

in order to solve for parcels using parks in adjacent counties. If the input counties are of a higher
complexity than Wake County, using the focus area approach described above will aid in
improving the completion rates.
5.5 Future Applications
The motivation for this study stems from the role that planners have in park and recreation
planning from site selection, funding allocation, and maintenance and operation. Having a
method such as the one proposed can greatly reduce the generalization that is often present in
current park system analysis. Wake County is a suburban community with very fragmented
urban areas. This study shows how the difference between the ODMT and a buffer based
coverage approach has a high degree of separation. However, in the past, funding and site
selection studies utilize buffer-based coverages to determine the priority areas.
As shown in the comparison of the ODMT to the buffer coverage, the issues of access
paths, appropriately located entry points, distribution of services and land area, and user demand
are not readily discernible in the buffer; this creates a highly generalized view of the park
system. Using the ArcGIS Network Analyst to conduct this cadastral-based workflow creates an
output that incorporates these elements and creates a grounded analysis revealing significant
gaps in the areas of the county depending on community.
Funding, such as that from the Parks and Recreation Trust Fund (PARTF), awards on a
need-based calculation that evaluates the number of people served as well as income and
demographics. A challenge facing this program is ensuring that not only are the intended
populations being served by the facility, but also verifying that the funds are being equally
distributed across all of the communities. PARTF looks at two variables in their ranking matrix,
location of residents and location of neighboring commercial services. Dense urban areas will

107

score higher when using a coverage-based approach compared to more suburban developments.
The census data for the urban areas is highly fragmented into a large number of objects;
observable when looking at Cary, NC compared to Raleigh, NC. Utilizing the cadastral
approach will eliminate the potential bias of the coverage approach. The cadastral approach also
aids planners in quickly mapping the residents within the 1/4 mileservice area for targeted
surveys.
In a reverse of this workflow, the cadastral approach will also allow for the conducting of
user surveys at each facility location. Collecting the user information over a 30-day or 45-day
period will facilitate executing an ODMT solution that maps the address of the users to the park
location, creating an actual user group based service area for each facility within the system.
This would show the average travel distances and frequency of visits per park, that when linked
back to the overall workflow, would allow further ranking of the parks based on the user count
and distances traveled. This would be of benefit when evaluating the effectiveness of a facility
in terms of total users and real service area. This will also allow making decisions about the
value of a specific park within the system, highlighting either the need for improvements to that
facility or the potential for selling the facility in exchange for a more beneficial site.
5.6 Final Conclusions
While automating the proposed workflow is not complete, the successful completion of the test
process utilizing a manual approach to the calculations shows that there can be benefit to
focusing on the parcel-to-parcel approach of measurement. Knowing which pathways are
traversable for a resident accessing a specific park parks and being able to measure these travel
distances accurately through the ODMT is a valuable component in reviewing the long-range
needs of a park system. This type of process can potentially be reverse integrated into a site

108

selection tool to allow planners to see how well a particular site will suit the community
residents.
The limitations of the computer hardware and file size will continue to improve and as
computational resources, increase further automations to the proposed workflow will be possible.
The continual evolution of computer hardware in conjunction with GIS data storage and
acquisition aids significantly in evaluating and planning for the future needs of a community in
an equitable and equally accessible manner. However, the costs may be outweighed by the
advantages of a ground based measurement technique compared to those of generalized areal
units, like containers and buffers, and may prove cost effective in identifying more appropriate
land acquisitions and even identifying and prioritizing surplus properties that could be used for
park development.
The overall benefits of the cadastral-based technique fill the voids found in literature by
introducing a means that combines the strengths of network analysis methodologies to proven
dasymetric techniques for apportioning the population data to the mapped dwelling locations.
This type of approach will help bring better validity in future uniformity of access assessments
and serve as a solid foundation for equality assessments.

109

REFERENCES
Boone, Christopher G., Geoffrey Buckley, J. Morgan Grove, and Chona Sister. 2009. “Parks and
people: An Environmental Justice Inquiry in Baltimore, Maryland.” Annals of the
Association of American Geographers 99: 767-87. Accessed August 14, 2014.
http://www.researchgate.net/publication/238043978_Parks_and_People_
An_Environmental_Justice_Inquiry_in_Baltimore_Maryland.
Carleyolsen, Sanya, Tanya Meyer, Joseph Rude, Ian Scott. 2005. "Measuring the Economic
Impact and Value of Parks, Trails, and Open Space in Jefferson County: Accounting for
Current and Future Scenario." Paper presented at the Urban and Regional Planning
Workshop Fall 2005, University of Wisconsin-Madison, Madison, Wisconsin, December
15, 2005.
Chandio, Imtiaz Ahmed, Abd Nasir B. Matori, and Dano Umar Lawal. 2011. “GIS-Based
Accessibility Analysis Using Suitable Land for Public Parks in Larkana City Pakistan.”
Research Journal of Applied Sciences, Engineering and Technology 3: 553-57. Accessed
August 1, 2014. http://www.researchgate.net/publication/233883820_GIS-
based_Accessibility_Analysis_Using_Suitable_Landfor_Public_Parksin_Larkana_City_P
akistan.
City of Raleigh (Raleigh). 2014. "Raleigh Demographics." City of Raleigh Business and
Development Department, September 23. Accessed September 29, 2014.
http://www.raleighnc.gov/government/content/PlanDev/Articles/LongRange/RaleighDem
ographics.html

110

Dalton, Alice M., Andrew Jones, David Ogilvie, Mark Petticrew, Martin White, and Steven
Cummins. 2013. "Using Spatial Equity Analysis in the Process Evaluation of
Environmental Interventions to Tackle Obesity: the Healthy Towns Programme in
England." International Journal for Equity in Health 12: 43-53. Accessed July 2, 2014.
http://www.equityhealthj.com/content/12/1/43.
Duncan, Dustin T., Ichiro Kawachi, Kellee White, and David R. Williams. 2012. "The
Geography of Recreational Open Space: Influence of Neighborhood Racial Composition
and Neighborhood Poverty." Journal of Urban Health: Bulletin of the New York
Academy of Medicine 90: 618-631.
Esri. 2014. “Network Analyst Tutorial.” ArcGIS Resource Center, July 1, 2012. Accessed July 1,
2014. http://help.arcgis.com/en/arcgisdesktop/10.0/pdf/network-analyst-tutorial.pdf
Hass, Kara. 2009. “Measuring Accessibility of Regional Parks: A Comparison of Three GIS
Techniques.” Master’s thesis, San Jose State University. Accessed August 14, 2014.
http://scholarworks.sjsu.edu/cgi/viewcontent.cgi?article=4637&context=etd_theses.
Heckert, Megan. 2013. “Access and Equity in Greenspace Provision: A Comparison of Methods
to Assess the Impacts of Greening Vacant Land.” Transactions in GIS. 17: 808-27.
Accessed September 5, 2014.
http://onlinelibrary.wiley.com.libproxy.usc.edu/doi/10.1111/tgis.12000/abstract
Higgs, Gary, Richard Fry, and Mitchel Langford. 2012. “Investigating the Implications of Using
Alternative GIS-based Techniques to Measure Accessibility to Green Space.”
Environment and Planning B: Planning and Design 39: 326-43. Accessed September 5,
2015. http://www.envplan.com.libproxy.usc.edu/epb/fulltext/
b39/b37130.pdf

111

James B. Holt, C.P. Lo, and Thomas W. Hodler. 2004. “Dasymetric Estimation of Population
Density and Areal Interpolation of Census Data.” Cartography and Geographic
Information Science 31: 103-121. Accessed September 5, 2014.
http://www.researchgate.net/publication/250016044_Dasymetric_Estimation_of_Populati
on_Density_and_Areal_Interpolation_of_Census_Data.
Joseph, Laura D. 2011. "Park Accessibility in Atlanta." Master’s thesis, Georgia State
University. Accessed July 27, 2014. http://scholarworks.gsu.edu/geosciences_theses
/32Langford.
Langford, Mitchel. 2007. “Rapid Facilitation of Dasymetric-based Population Interpolation by
Means of Raster Pixel Maps.” Computers, Environment and Urban Systems 31: 19-32.
Accessed September 5, 2015.
http://www.sciencedirect.com/science/article/pii/S0198971506000330
Lindsey, G., Maltie Maraj, and Son Cheong Kuan. 2001. “Access, Equity, and Urban
Greenways: An Exploratory Investigation.” The Professional Geographer 53: 332-46.
Accessed September 6, 2014. http://onlinelibrary.wiley.com/doi/10.1111/0033-
0124.00288/pdf
Liu, Lin, and Xia Li. 2014. “GIS and Built Environment.” Geojournal 79: 127-8. Accessed
August 1, 2014. http://link.springer.com/article/10.1007/s10708-013-9507-9.
Loukaitou-Sideris, Anastasia and Orit Stieglitz. 2002. "Children in Los Angeles Parks: A Study
of Equity, Quality and Children's Satisfaction with Neighborhood Parks." The Town
Planning Review 73: 467-488. Accessed August 14, 2013.
http://www.jstor.org/stable/40112531

112

Lynch, Kevin. 1960. The Image of the City. Cambridge, MA: MIT Press.
Lwin, Ko, and Yuji Murayama. 2012. “GIS Network Model in Geospatial Analysis.” In Progress
in Geospatial Analysis, edited by Yuji Murayama, 183-194. Tokyo: Springer Japan.
Accessed August 5, 2014. http://link.springer.com/chapter/10.1007%2F978-4-431-
54000-7_12
Maantay, Juliana Astrud, Andrew R. Maroko, and Christopher Herrmann. 2007. “Mapping
Population Distribution in the Urban Environment: The Cadastral-Based Expert
Dasymetric System (CEDS).” Cartography and Geographic Information Science 34: 77-
102. Accessed September 5, 2014. http://www.lehman.edu/deannss/
geography/publications/Dasymetric_CaGIS_Maantay.pdf
Makse, Hernán A., Shlomo Havlin, and H. Eugene Stanley. 1995. “Modelling Urban Growth
Patterns.” Nature 377: 608-12. Accessed July 20, 2014.
http://www.nature.com/nature/journal/v377/n6550/pdf/377608a0.pdf
Maroko, Andrew R., Juliana A. Maantay, Nancy L. Sohler, Kristen L. Grady, and Peter S. Arno.
2009. “The Complexities of Measuring Access to Parks and Physical Activity Sites in
New York City: A Quantitative and Qualitative Approach.” International Journal of
Health Geographics 8: 34-57. Accessed July 5, 2014.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2708147/pdf/1476-072X-8-34.pdf
Marusic, Barbara Golicnik. 2011. “Analysis of Patterns of Spatial Occupancy in Urban Open
Space Using Behaviour Maps and GIS.” URBAN DESIGN International 16: 36-50.
Accessed August 7, 2015. http://www.palgrave-
journals.com/udi/journal/v16/n1/full/udi201020a.html

113

Mertes, James D. and James R. Hall. 1996. Park, Recreation, Open Space and Greenway
Guidelines. Arlington, VA: National Recreation and Park Association.
Moeller, John. 2014. “Standards for Outdoor Recreational Areas.” American Planning
Association. Accessed September 5, 2014. http://www.planning.org/pas/at60/
report194.htm.
Nicholls, Sarah. 2001. “Measuring the Accessibility and Equity of Public Parks: A Case Study
Using GIS.” Managing Leisure 6: 201-19. Accessed August 10, 2014.
http://staff.washington.edu/kwolf/Archive/Classes/ESRM304_SocSci/304%20Soc%20Sc
i%20Lab%20Articles/Nicholls_2001.pdf
Office of State Budget and Management (OSBM). 2014. “2013 Certified County Population
Estimates.” Office of State Budget and Management, September 19. Accessed September
25, 2014. http://www.osbm.state.nc.us/ncosbm/facts_and_figures/socioeconomic_data/
population_estimates/demog/countygrowth_cert_2013.html.
Sister, Chona, Jennifer Wolch, and John Wilson. 2010. “Got Green? Addressing Environmental
Justice in Park Provision.” GeoJournal 75: 229-248. Accessed September 5, 2014.
http://link.springer.com/article/10.1007%2Fs10708-009-9303-8.
Talen, Emily, and Luc Anselin. 1998. “Assessing Spatial Equity: An Evaluation of Measures of
Accessibility to Public Play-grounds.” Environment and Planning A, 30: 595-613.
Tarrant, Michael, and H. Ken Cordell. 1999. "Environmental Justice and the Spatial Distribution
of Outdoor Recreation Sites: An Application of Geographic Information Systems."
Journal of Leisure Research 31:18-34. Accessed August 8, 2013.
http://search.proquest.com.libproxy.usc.edu/docview/201178702/fulltextPDF/1CEF0A8C
1A56424CPQ/1?accountid=14749.

114

Town of Cary (Cary). 2012. “Chapter 4: Inventory & Level of Service Analysis.” Parks and
Open Space Master Plan. Cary, NC: Town of Cary, November 13. Accessed July 30,
2014. http://www.townofcary.org/Departments/fdts/facilities/Current_Projects
/prcrmasterplan.htm.
Wake County GIS Department (WakeA). 2014. Wake_Property_2014_06. Wake County, NC:
Wake County GIS Department, June 30. Accessed July 1, 2014.
http://www.wakegov.com/gis/Pages/default.aspx.
Wake County GIS Department (WakeB). 2014. Wake_Census_2010. Wake County, NC: Wake
County GIS Department, June 30. Accessed July 1, 2014.
http://www.wakegov.com/gis/Pages/default.aspx.
Wake County GIS Department (WakeC). 2014. Wake_PublicOpenSpace_2014_06. Wake
County, NC: Wake County GIS Department, June 30. Accessed July 1, 2014.
http://www.wakegov.com/gis/Pages/default.aspx.
Wake County GIS Department (WakeD). 2014. Wake_Streets_2014_06. Wake County, NC:
Wake County GIS Department, September 5.
http://www.wakegov.com/gis/Pages/default.aspx. Accessed July 1, 2014.
Wake County GIS Department (WakeE). 2014. WakeMajorRds. Wake County, NC: Wake
County GIS Department, September 5. Accessed July 1, 2014.
http://www.wakegov.com/gis/Pages/default.aspx.
Wake County Planning Department (Wake). 2014. Unified Development Ordinance of Wake
County. Wake County, NC: Wake County Planning Department, October 6. Accessed
November 1, 2014. http://www.wakegov.com/planning/zoning/Documents
/adoptedordinance.pdf

115

Wolch, Jennifer, John P. Wilson, and Jed Fehrenbach. 2005. "Parks and Park Funding in Los
Angeles: An Equity-mapping Analysis." Urban Geography 26:4-35. Accessed
August 5, 2014. http://www.tandfonline.com/doi/pdf/10.2747/
0272-3638.26.1.4#.VMVDMrt0wiQ.

Abstract (if available)

Abstract Park planners make long-term land acquisition and capital improvement plans based in part on population growth and gap analysis of existing facilities. This study demonstrates a new cadastral-based technique to measure park access for residents in Wake County, NC. Based on road network and cadastral data, the technique uses the Origin-to-Destination Matrix Tool within Esri’s Network Analyst extension in conjunction with dasymetric mapping of US Census Data to the cadastral data. The demonstrated workflow provides for a highly detailed assessment of walking distance between parcels and parks, that when linked with the population data, provides a gap analysis based on the amount of parkland and number of parks available at each parcel. Successful completion of an analysis at this level of detail illustrates a very different view of park coverage for Wake County, NC compared to traditional methods, revealing how hard edges created by major thoroughfares and soft edges created by property ownership impact pedestrian accessibility. Using the cadastral-based method, 19.85% fewer parcels have 1/4-mile park access than compared to a buffer based method (6.72% versus 26.27%). The use of this type of technique will allow for a more comprehensive assessment of the peoples served by the park system and when coupled with demographic information, may prove more effective in assessing grants and monitoring the impact of public initiatives promoting equality and uniformity of access to public parks.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Mapping future population impacts caused by sea level rise in Huntington Beach and Newport Beach: comparing the cadastral-based dasymetric system to past dasymetric mapping methods

PDF

Analysis of park accessibility in Redan, Georgia Web GIS application

PDF

The role of amenities in measuring park accessibility: a case study of Downey, California

PDF

Modeling patient access to point-of-care diagnostic resources in a healthcare small-world network in rural Isaan, Thailand

PDF

Estimating populations at risk in data-poor environments: a geographically disaggregated analysis of Boko Haram terrorism 2009-2014

PDF

Network accessibility and population change: historical analysis of transportation in Tennessee, 1830-2010

PDF

Generating bicyclist counts using volunteered and professional geographic information through a mobile application

PDF

Campaign financing for the U.S. House of Representatives: an interactive Web map

PDF

A fire insurance map geocoder for pre-earthquake San Francisco

PDF

Address points and a maser address file: improving efficiency in the city of Chino

PDF

Object detection and digitization from aerial imagery using neural networks

PDF

Measuring seasonal variation in food access: a case study of Everett, Washington

PDF

Access to active play parks for youth segments in Alexandria, Virginia

PDF

Using volunteered geographic information to model blue whale foraging habitat, Southern California Bight

PDF

Investigating electoral college reform: geography's impact on elections, and how maps influence our perception of election outcomes

PDF

Social media to locate urban displacement: assessing the risk of displacement using volunteered geographic information in the city of Los Angeles

PDF

Integrating land survey data into measurement-based GIS: an assessment of challenges and practical solutions for surveyors in Texas

PDF

Analyzing earthquake casualty risk at census block level: a case study in the Lexington Central Business District, Kentucky

PDF

Guiding business oriented volunteered geographic information through geotigger services: a case study of CrossFit affiliates

PDF

Spatiotemporal visualization and analysis as a policy support tool: a case study of the economic geography of tobacco farming in the Philippines

Asset Metadata

Creator Parsons, Jonathan Edward (author)

Core Title Mapping uniformity of park access using cadastral data within Network Analyst in Wake County, NC

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 02/11/2015

Defense Date 01/12/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag accessibility,cadastral mapping,dasymetric mapping,GIS,Network Analyst,OAI-PMH Harvest,origin-to-destination pairs,Parks,population density,service areas

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Vos, Robert O. (committee chair), Chiang, Yao-Yi (committee member), Warshawsky, Daniel N. (committee member)

Creator Email jeparson@usc.edu,jparsonsRLA@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-532459

Unique identifier UC11298302

Identifier etd-ParsonsJon-3184.pdf (filename),usctheses-c3-532459 (legacy record id)

Legacy Identifier etd-ParsonsJon-3184.pdf

Dmrecord 532459

Document Type Thesis

Format application/pdf (imt)

Rights Parsons, Jonathan Edward

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA