Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Walkability study for school accessibility: case study of the San Juan, Puerto Rico elementary schools
(USC Thesis Other)
Walkability study for school accessibility: case study of the San Juan, Puerto Rico elementary schools
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Walkability Study for School Accessibility:
Case Study of the San Juan, Puerto Rico Elementary Schools
by
Joel Rodríguez
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)
December 2015
Copyright 2015 Joel Rodríguez
ii
DEDICATION
I dedicate this document to my grandmother Berta Camacho, my friends Suzika Pagán, Shannon
Prescott and Heather Kelley for their constant support, patience and understanding. Also I want
to recognize my committee chair Dr. Robert Vos, who has helped me through this project in
understanding the concepts and theories behind the research and for his support in completing
this project.
iii
ACKNOWLEDGMENTS
I will be forever grateful to my mentor, Dr. Robert Vos. Thank you for your patience and
understanding. Tenacity and persistence are qualities I have come to appreciate through this
process.
iv
TABLE OF CONTENTS
DEDICATION ii
ACKNOWLEDGMENTS iii
LIST OF TABLES vii
LIST OF FIGURES viii
LIST OF ABBREVIATIONS x
ABSTRACT xi
CHAPTER 1: INTRODUCTION 1
1.1 Defining Accessibility 3
1.2 Service Area to Walkability 3
1.3 Measuring Population 4
1.4 Case Study Area 5
1.5 Motivation 7
CHAPTER 2: BACKGROUND AND LITERATURE REVIEW 10
2.1 Pedestrian Access to Facilities 11
2.2 Walkability and Pedestrian Access to School 12
2.3 Dasymetric Mapping 17
2.4 Summary 18
CHAPTER 3: METHODOLOGY 19
3.1. Data Selection and Output Products 21
3.2 Population per Parcel 24
3.2.1 Population per Parcel Calculation 25
v
3.2.2 Dasymetric Mapping 26
3.4 Walkability 29
3.4.1 Accessible Roadways 29
3.4.2 Service Area Methodology 33
3.4.3 Walkability Output 33
3.5 Student Population Affected 37
3.6 Origin Destination Cost Matrix Analysis 38
CHAPTER 4: RESULTS 40
4.1 Network Analysis Issues and Solutions 40
4.2 School Accessibility Results 41
4.3 Environmental and Urban Obstacles 49
4.3. Changes in Average Distance 50
4.4 Scenario Analysis 56
CHAPTER 5: DISCUSSION AND CONCLUSIONS 61
5.1 Benefits of Using Network Analyst and OD Cost Matrix 61
5.1.1 Accessibility Measures 62
5.1.2 Service Area Analysis 63
5.1.3 OD Analysis 64
5.2 Limitations of the Project 64
5.3 Future Research 66
5.4 Final Conclusion 67
vi
References 68
vii
LIST OF TABLES
Table 1 Source Data Output 24
viii
LIST OF FIGURES
Figure 1 San Juan County Schools ................................................................................................. 6
Figure 2 Overview diagram of study workflow ............................................................................ 21
Figure 3 ModelBuilder Showing Selection of Parcel and Spatial Join ......................................... 26
Figure 4 Student Population per Parcel Calculated ...................................................................... 27
Figure 5 Student Population per Parcel Distribution .................................................................... 28
Figure 6 Walkable Roadways in San Juan County ....................................................................... 32
Figure 7 ModelBuilder Showing Selection of Parcel and Spatial Join ......................................... 33
Figure 8 Walkability Area before School Closure ........................................................................ 35
Figure 9 Walkability Areas after Closure ..................................................................................... 36
Figure 10 Parcels Affected Inside Service Area ........................................................................... 37
Figure 11 OD Cost Matrix Analysis of a School Facility ............................................................. 38
Figure 12 Graph of 1/4 Mile Access Results ................................................................................ 42
Figure 13 Student Population Accessibility before Closure ......................................................... 43
Figure 14 Student Population Accessibility after Closure ............................................................ 44
Figure 15 Schools in San Juan Most Affected School by Closure ............................................... 45
Figure 16 Student Population at a 1/4 Mile Access before School Closure ................................. 47
Figure 17 Student Population at a 1/4 Mile Access after School Closure .................................... 48
Figure 18 Laguna San José and Laguna Los Corozos .................................................................. 49
Figure 19 Mean Distance Walked Before and after School Closure ............................................ 51
Figure 20 Mean Distance Walked Before School Closure ........................................................... 53
Figure 21 Mean Distance Walked After School Closure.............................................................. 54
Figure 22 Histogram of Total Distance Walked Before School Closure...................................... 55
ix
Figure 23 Histogram of Total Distance Walked After School Closure ........................................ 56
Figure 24 Case Scenario of School Closure at ¼ mile access ...................................................... 58
Figure 25 Bar Graph 1/4 Mile Access to school before and after school closure scenario .......... 59
Figure 26 Bar Graph Total Distance Walked to School Scenario ............................................... 60
x
LIST OF ABBREVIATIONS
CRIM El Centro de Recaudación de Ingresos Municipales
CDC Center for Disease Control
DE Departamento de Educación
DTOP Departamento de Transportación y Obras Públicas
FHWA Federal Highway Administration
GIS Geographic Information Systems
GIST Geographic Information Science and Technology
MAUP Modifiable Areal Unit Problem
PRTE Reorganization and Transformation School program
SSI Spatial Sciences Institute
US United States
USC University of Southern California
xi
ABSTRACT
Geographic Information Systems (GIS) mapping applications have proven to be an integral part
in school site planning. However, most school site planning does not take walkability into
account. This study describes a method to measure how walking access to schools was affected
by the closure of elementary schools in the San Juan County. Recent studies in students walking
to school in the US have found that there has been a major decrease overall (National Household
Travel Society 2013). Using population density and dasymetric mapping, the number of students
in each parcel in the San Juan School County was estimated. A walkability service area was
derived from a network dataset using the functional classification of roadways from the Federal
Highway Administration (FHWA). Accessibility was calculated ¼ miles away from the school
using the service areas before and after school closure. It was determined that after school
closure pedestrian accessibility and total distance walked to school did not change significantly.
Network analysis represents a direct approach that assesses accessibility and physical barriers of
the urban environment. Combining the walkability service area and the population per parcel,
student accessibility was calculated. The use of this methodology will allow a better assessment
for the school site planning and can be used to develop initiatives that will promote walking to
and from school.
1
CHAPTER 1: INTRODUCTION
Geographic information systems (GIS) are designed to capture, manipulate, analyze, manage,
store and display geographical data, and they are an important tool for planning and development
of spatial events (Longley, et al. 2011). As time goes by and technology evolves, GIS mapping
technology can be implemented in many aspects of our society. Carey (2011) explains that GIS
mapping is an important process for projecting student enrollments for school locations, which is
an integral part for school site planning. Carey (2011) further explains that with GIS mapping
technology, planners and researchers have the ability to incorporate factors that affect school
planning such as race, school enrollment, proximity of schools to demographic groups, growth
and decline of neighborhoods. All of these factors can be considered for overall impacts on
school districts and siting of school locations (Carey 2011) (K. D. Carey 2011). Carey also points
out the importance of mapping in school planning, and how walking as a mode of transportation
to schools is an effective and economic practice that should be implemented. The US
Environmental Protection Agency (EPA) has addressed the importance of the site location of a
school, accessibility, and walking to and from are school, and how they greatly affect the
communities around the school (EPA 2013). Walking to and from school is usually referred to
as walkability, and it should play an important role in school site location.
Although there are numerous definitions for walkability, it is defined here as a measure of
how accessible an area is to walking (Bliesner, Bouton and Schults 2010). Walkability is the
combination of the physical activity of walking and the physical elements that support it, like the
presence of sidewalks, which make up the environment that can be walked on. Walkability
should be considered an important measure of how active and suitable a city is, because it
represents how the populations of certain areas move to and from one place to another. The
2
intention of this study is to conduct a walkability assessment to see the population affected
following the elementary school closure in the San Juan, Puerto Rico County. It is the intention
of this study that walkability could be then added to the criteria considered in making decisions
about which schools to keep and which school to close. Currently, the decision to close the
school was made without taking walkability into account, seeing that 65 percent of the student
population walks to and from school (Departament de Educacion de Puerto Rico 2014).
Using population density, roadways, and parcel data along with network analysis in GIS, a
walkability analysis can be created, displaying which schools are easily walked to and how
school closures will affect them. To develop a base for this study, the ArcGIS Network Analyst
extension was researched to better understand the use of a road network in a spatial analysis.
Accessibility was defined using the roadways classifications. This classification was based on
the functional classification of roads from the Federal Highway Administration (FHWA). This
organization classifies roadway groups in streets and highways according to the service they
provide.
Once the accessible routes were identified, ArcMap software can be used to calculate
walking distance between the schools and residential parcels using the network analyst tool and
then Census data was used to calculate population per household. This information was used to
determine how the closing of schools affected the population of the school in the San Juan
County.
3
1.1 Defining Accessibility
Accessibility to schools is determined by the capability of a person to reach a destination. This
can be usually answered using a car or as conceived in this study simply walking. How
accessibility is defined for schools influences greatly the population of students affected by the
schools closure. Not all places have the same accessibility. Thus, the way this study measures
accessibility is important for construction of the service area that represent the walkability area to
schools. There are two main concepts that accessibility relies on: location and distance. Location
is the relative space in relation to the transport infrastructure. The second concept of accessibility
is distance, which is the physical separation between locations (Rodrigue 2013). This can only
exist when there is a separation between locations. In the case of this study, there is a separation
between the schools and the parcels where students reside.
For this study the approach that followed involved the definitions of the roadways as
presented by the FHWA, which are applicable to roadways in Puerto Rico. Even though
roadways are mainly associated with vehicle transportation, they serve as the main structure for
walking. Not all roadways are meant to be walkable but using the FHWA classification of
roadways it is possible to identify which roadways are walkable and which one are not. This
approach also does not take into account temporary physical barriers that prevent people from
walking like road construction. While accessible deals more with the network created by the
roadways, walkability deals with the service area around the schools. Accessible roadways are
accessible by walking.
1.2 Service Area to Walkability
Service areas in GIS are determined from a network of roads and it determines the distance
around a location that a person travels on a network. How service areas are defined for a study
4
greatly influences the population affected by the closure of the schools. Some methods to
calculate service area involve the use of buffers around the locations to travel (Yin and Zhan
2014). The problem with the use of buffers is that this method does not take into account
roadways, buildings, topography or any obstruction around the schools. In truth, buffer analysis
poorly represents accessibility to any location.
A main concern with buffers is that parcels that fall inside the buffer area may not be able
to access the schools because buffer areas do not represent the true distance along the road
network from the parcels to the schools. Whereas, service area analyses does represent a better
spatial analysis for accessibility to and from schools. Service areas describe a polygon around the
facility selected based on a network dataset created from the roads. In addition to roads, this
network can contain stop signs, street signs, and other things that affect walkability can be added
to the network dataset to increase the service area analyses accuracy. Also, service area analysis
can be set up in a way that they will not overlap with other facilities. This can be done in case
some of the schools are too close together as a way of estimating what neighborhoods should
send students to particular schools.
1.3 Measuring Population
One of the reasons to create a service area is to assess the student population that has access to
schools. The source for population data comes from the United States (US) Census and was
further edited by the Sistemas de Datos Geograficos (Portal Datos Geograficos Gubernamentales
2003). By using the smallest scale of measurement for the census block it is possible to calculate
the parcel around the service areas of the schools. This assessment would not be possible using
buffer analysis because in a buffer analysis some parcels are not directly ¼ miles always from
5
the schools. Alongside the census information and the parcel data type, it is possible to calculate
the student population in each parcel.
Dasymetric mapping is method of thematic mapping that utilizes areal symbols to
spatially classify data (Mennis 2003). When dealing with census and parcel data it is possible to
calculate the actual population in each parcel based on a shared attribute. This proved to be an
accurate assessment of the student affected by the closure of the elementary school in the San
Juan County.
1.4 Case Study Area
Puerto Rico is located in the Caribbean Sea to the east of the Dominican Republic and north of
South America. The island extends 115 miles from east to west and 41.66 miles from north to
south making it the smallest island in the Greater Antilles. As an unincorporated territory of the
US, Puerto Rico is officially known as the Commonwealth of Puerto Rico and it comprises 78
counties with San Juan being the main capital. An initial report from El Departamento de
Educación (Puerto Rico Department of Education, DE) indicates that 73 schools (Figure 1) were
closed between 2013 and 2015 (Departament de Educacion de Puerto Rico 2014). Not only San
Juan County is the most populated county on the island, but it is also the county most affected by
the school closure (Instituto de Estadísticas de Puerto Rico 2014).
6
Figure 1 San Juan County Schools
7
1.5 Motivation
School construction projects such as renovations and new buildings are occurring continually in
Puerto Rico. Decisions about construction greatly affect and influence communities. It is not
only communities, but the schools inside of them that face changes in their environmental
atmospheres and economies. These changes are usually overlooked by the government and
sometimes their effects go unnoticed until the damage is already done. An Environmental
Protection Agency (EPA) study finds that school proximity matters, meaning that students with
shorter times to walk to or from school are more likely to walk (EPA 2013). Also, the study
determined that it is the built environment that influences the options of traveling to school
because students traveling through higher quality and more accessible roads are more likely to
walk. As a result of travel behavior difference, school locations have an impact on air emissions.
If more accessibly located schools can be reached by walking, then this can help with the
reduction of air pollution (EPA 2013). These results suggest that smart growth planning,
accomplished in part through GIS mapping can act as a tool to improve students walking
behavior.
Others studies have found a significant decrease in the numbers of children in the US
walking and biking to school. From 2000 to 2010, the National Household Travel Survey
(NHTS) reported that 15% of students between 5 and 15 years old walked to and from school,
and 1% biked (National Household Travel Society 2013). An older study from the Federal
Highway Association (FHWA) found that for the years 1969 to 1990, 48% of students walked or
biked to school (FHWA 1972). This is an alarming decrease that should not be taken lightly. In a
survey by the Center for Disease Control (CDC), it was found that this decline in walking to and
8
from school was mostly due to accessibility and distances traveled by the students to walk to
school (Dellinger and Staunton 1999).
Several organizations in the US have started to pay attention to the decrease in students
walking to school and are trying to make a difference by implementing new programs to increase
the quality of walking and accessibility to school. Institutions like the “Safest Routes to School
Programs” within the State of California, works to increase the number of children who walk or
bike to school by funding projects and programs that remove the barriers that currently prevent
them from doing so (Dellinger and Staunton 1999). Collected data that the organization
considers walking barriers included lack of infrastructure and unsafe infrastructure, such as
construction sites and cul-de-sac roadways. This data is then routed to other organizations and
agencies like the National Center for Safe Routes to School and the federal government. These
organizations use these data to get rid of these barriers. Even the US Department of Health and
Human Services, and the CDC have started a Kids-Walk-to-School campaign, citing the rising
rates of childhood obesity, diabetes, and asthma as reasons to encourage walking to school
(Centers for Disease Control and Prevention 2011).
1.6 Problem Statement
In November of 2014, the DE in a project called “Plan de Transformación y Reorganización de
Escuelas” (Reorganization and Transformation School program, PRTE), decided to close 73
schools, and announced future plans to close 580 more by 2020 (Departament de Educacion de
Puerto Rico 2014). After a study performed by The Boston Consulting Group (BCG), a
consulting firm on business strategy, they discovered that the enrollment in public schools went
down by 41% from 1980 to 2013 (Departament de Educacion de Puerto Rico 2014).
9
The BCG study and DE action did not take into account how school closures could affect
the communities around the closed schools and those students that walked daily to and from
school. The DE criteria for choosing which school to close included: physical condition of the
school, academic performance, utilization, closeness to a similar school, and student enrollment
(Departament de Educacion de Puerto Rico 2014). Proximity to a similar school may function as
a sort of proxy criteria for maintaining walkability, but this deserves measurement and testing.
The use of GIS technology for smart school site planning, accessibility and walkability
can be used to better understand this issue and even find a solution. This research attempts to add
to the literature and propose a methodology to measure accessibility and the impacts of the 73
schools closing to the communities’ surroundings in the San Juan County, Puerto Rico school
district.
10
CHAPTER 2: BACKGROUND AND LITERATURE REVIEW
For many years people did not rely on vehicle transportation to get to school. Walking to and
from school was part of our daily lives. As years go by and people became dependent on
automobiles for transportation this culture began to change around the world. It is our choice to
walk as a mode of transportation and this is a simple fact that defines any walkability study.
There are many elements involving walkability: population, land use, tree canopies, number of
intersections, accessibility, and steepness of terrain and so on. Many studies only focus their
efforts on defining walkability based on the elements found in their specific urban environment
(Chen 2012; Parson 2015; Yin 2014).
Urban sprawl often creates obstructions that hinder our journey from one place to another
(Speck 2013). Instead of deciding to walk, people start finding that it is a burden to get
anywhere. Walkability does not have to be a burden, but it can instead be a measure of a healthy
urban environment. There is no perfect combination of elements that truly defines walkability
and in some cases it goes beyond the urban environment into the needs of current society.
Populated and urban places usually are in need of walkable pathways. It is in populated places
where roadways are usually cramped with cars and traffic where the environment becomes
unbearable (Kunstler 1994). Populated places also have many types of locations that are closer
together. This means that it should be easy to walk from one place to another. Less populated
and rural places do not have the same need for walkable pathways. Locations are far apart and
roads are not very walkable. The San Juan County is an urban and populated place with many
roadways and walkable pathways. At the time of this study, no type of walkability study in
Puerto Rico has been published to the best of the author’s knowledge. This fact makes this study
new and unique.
11
The purpose of this chapter is to explore the literature and compare the different methods
used to measure walkability. How have previous studies assessed and measured accessibility?
Approaches to accessibility vary depending on the area and focus of the study. This was taken
into account for this research. Defining accessibility in this study takes taken a different
approach from some of the studies discussed in this section. The literature review also discusses
some methods used to measure walkability. Most previous studies on walkability reviewed used
a buffer analysis approach with Census data organized by Census geographical units. However,
this approach does not work well in urban or suburban environments where roads have no direct
access to the locations and there are urban and environmental barriers. Also, as an alternative to
relying solely on Census geographical units, dasymetric mapping is discussed in this section. The
following discussion establishes the basis for this research and sheds more light on the optimal
elements to include in walkability metrics.
This chapter is organized into four subsections: accessibility, walkability, and dasymetric
mapping of student population. A thorough understanding of these concepts influences the
makeup of this study. The foundation of this study relies on bringing these concepts together to
interpret and assess a spatial event. At the end of this chapter, a summary is provided that
outlines how the main findings build the method proposed for this study.
2.1 Pedestrian Access to Facilities
Parson’s (2015) research describes the use of cadastral-based applications and buffer mapping
techniques to measure access to parks for residents. Cadastral-based applications utilize
dasymetric mapping to estimate the number of people that live in each parcel. This way it is
possible to measure how many people have access to parks. Using road network data and
cadastral data, along with Network Analyst extension in ArcMap, Parson’s (2015) produced a
12
walking distance assessment between parcels and parks in Wake County, North Carolina. By
comparing the traditional buffer method with the cadastral-based method, he showed that the
cadastral-based method was more accurate than the buffer method. The buffer method covers
areas around parks without regards to accessibility, unlike the cadastral method that allowed him
to see which parks are more accessible using urban barriers and walkability. Parson’s spatial
analysis is built in a model to be used as a decision support.
Parson’s (2015) research demonstrated that the buffer methods do not always work and
that is a very unreliable method to analyze walkability. Buffers do not assess physical barriers in
urban environments and thus do not show true accessibility to areas. To measure the
accessibility to parks, Parsons (2015) utilized service areas analysis to derive a walkability path
to and from the parcel. This service area was derived from a network dataset created from the
roads. Once the service areas were derived, Parsons calculated the population on those service
areas.
Unlike Parsons (2015) who focused on parks, in this study the service areas represent an
area were the population has access to schools. The service area that he used was derived from a
network data dataset created from the roadways. For his study Parson’s (2015) did not utilize any
criteria on the roads. His network data set was based mainly on local roads and did not take into
account other walkable pathways. For this thesis, a similar approach was taken but the criteria
for the creation of the network data set was adapted to the context of the Puerto Rico road
network and school accessibility problem.
2.2 Walkability and Pedestrian Access to School
Typically, walkability has been measured by the use of buffer areas or service areas around the
target location. But to truly derive walkability areas, roadways have to be taken into account.
13
Chen (2012) has discussed how GIS can be used to contribute to bus network design by applying
the concept of walkability to bus route planning. She examined two bus routes in Orange County
California that were under different urban development patterns to determine if there is a
difference in the walkability score. The major two factors that influenced these urban
development patterns is the difference in street connectivity. These two urban set ups were
normal grid street and cul-de-sac streets. Normal Grid streets are traditional urban roads,
connected along arterials, highways and local streets, and cul-de-sac neighbors are usually dead
end roads that are meant to control flow of traffic. Using population density, street connectivity,
steepness and tree canopies she examined these different bus routes extracting a walkability
score for each. She concluded that normal grid streets are highly walkable compared to cul-de-
sac streets.
Chen’s (2012) thesis has the right idea of using population density and street connectivity
for her walkability index, and steepness and tree canopy are a great measurement of walkability
and could be important to measure the walkability in Puerto Rico. Factors that drastically affect
walkability usually deals with urban and environmental barriers. Even though she did not
conduct a full network analysis, street connectivity proved to be a major element in measuring
walkability. Chen (2012) also considered non-street paths like short cuts through parks. Without
street connectivity walkability is not possible. People that walk to and from place to places will
be deterred from walking if the pathways are intricate to walk. Normal grid street and cul-de-sac
street definitions can be viewed as elements of accessibility, normal grid street being more
accessible than cul-de-sac streets. This element plays a major role in walkability.
Yin’s (2014) research in walkability based in Wuhan, China utilized buffer analysis
approach. In his study, Yin (2014) integrates the factors of transportation, population and land
14
use to determine if the educational facilities in the Wuhan Province were in accessible places.
Yin (2014) highlights the need for a “green” mind set for Wuhan Province. In his view, for many
years urban development has been more oriented toward motor vehicle popularity without regard
for an active life style or a pollution free environment. The streets in the province have become a
harsh place to walk, and space for walking is slowly being swallowed up by non-accessible areas
used for motor vehicles. Pedestrians are being pushed back into less walkable pathways by the
rapid extension of urban sprawl. Places that normally used to be safe to walk have become more
vehicle friendly rather than pedestrian friendly.
From Wuhan City plans, Yin (2014) identified 2,294 educational facilities in the area. He
identified population density as the major factor in the measurement of walkability. He assumed
that if there is a high population density in an area, that area will have a larger and stronger
demand for walking, unlike areas with low-density population.
To evaluate walkability of the area, Yin (2014) added population density, road density,
and intersection density together in a raster format. He later compared the walkability index with
the land use to evaluate which areas were more or less walkable. He found that urban areas were
more walkable than rural areas. Yin’s (2014) research has some major deficiencies. He made too
many assumptions without regards of the built environment in Wuhan City. His buffers analyses
did not take into account that population does not have direct access to many types of locations.
Urban environment in this city are labyrinths of pathways that cannot be assessed by a buffer
area. Also his study relies too much on buffer analyses that did not cover urban barriers. There
needs to be a better understanding of the physical factors or elements in an urban environment
that affect walkability.
15
Leslie et al.’s (2006) research dealt with elements that were present and affected the
urban environment. His research was based on the idea that there were objective measures for
urban planning and transportation that help to better understand the relationship between factors
that define physical environment and physical behaviors of an urban environment. Coming from
this, she assumes that walking is the primary drive for environmental policies and public health.
To Leslie et al. (2006) there were elements built into the environment that may or may not
promoted walkability. GIS has the capability of using spatial data to create measure of
walkability. In this way, Leslie et al. (2006) believes that GIS can be used to create polices that
will encourage walking.
To Leslie et al. (2006) the choice of walking is greatly influenced by the built
environmental factors in a city. These elements are those that encourage walking, like shade,
steepness, and accessibility. One of the focuses of Leslie et al.’s (2006) research is based on the
role of land use in walkability. This comes from the idea that land use is a major influence for
walking. Where there is a high population and mixtures of land use, walking becomes more
practical (Butterworth and Edwards 2006). The more mixed the urban environment is, the shorter
are the distances in a city. Leslie (2006) also discussed the relationship between distance and
walking, scoring neighborhood paths. The longer the distance is in the urban environment the
less likely it is that someone will walk. Street connectivity also affects walkability. Where there
is better connectivity, the walking tends to be easier because fewer barriers are encountered.
Leslie et al.'s (2006) research described how four elements relevant for walking can be
used to create a walkability index. This index can be used to measure walkability to and from
schools, work and other destinations. These four elements were dwelling density, street
16
connectivity, land use mix, and net retail area. These elements were taken from raster and
network data.
According to Leslie et al. (2006) the concept of dwelling density deals, with the ratio of
number of dwelling units (parcel) to the land area in total. The level of street connectivity
emphasizes the use of street centerlines and number of connections. In addition, the ratio of the
number of intersections to the land area was calculated. Higher connectivity means greater
choices of potential routes and easier access to major roads and collectors. Land use mix
(accessibility and diversity) was also calculated by using the land use and zoning data. Using the
entropy score, which calculates the degree to which different land uses are scattered within an
area, provides a score with respect to the land use and land zone. The result of the score ranges
from 0 representing homogeneity (all lands uses are a single type) to 1, which represents
heterogeneity (developed areas are evenly distributes among all land use categories). This is
based on the idea that the more varied the land use is, the more conducive it is to walk to various
destinations.
Leslie (2006) standardized these four elements into four scores that can be summed in
order to obtain an overall walkability index. Derived from this, highly walkable areas were flat
with well-interconnected grid like streets, with higher population density and a mixture of
dwelling styles and high land use type. The low walkable areas were larger sized blocks with
fewer intersections and greater distance between them, low residential density, and homogenous
land use with little retail.
Based on what was learned from the studies described above, street connectivity plays an
important role into measure the network of the roadways for walkability. Roadways feed the
pedestrian from location to location and they support walkability.
17
2.3 Dasymetric Mapping
Holt (2011) researched the way spatial analysis was done when ancillary data, such as parcel
information is used. He saw that data collected and used by agencies usually is constrained by
the level of aggregation. This data when used for spatial analysis usually was problematic, and it
was difficult for the user to represent the spatial phenomenon without some sort of error. Not
only that, but this data also is also affected by the Modifiable Areal Unit Problem (MAUP). This
refers to the variation in scales and spatial aggregation of units that may results in a change in
statistical analysis results for areal data. When this type of data incompatibility is presented, it is
best to use a technique of areal interpolation. The purpose of Holt’s research was to describe one
technique of areal interpolation that can minimize MAUP, dasymetric mapping.
MAUP is defined as the occasion in which modifying the boundaries and scale of
aggregate data significantly affects the results of spatial data. This is a problem of great concern
when performing spatial analysis, especially when using census block or census tract data. It is
often not clear if results of census data display the reality of the region, a problem often seen in
Choropleth maps of population. Menni’s (2003) proposed surface model for representation of
population data as it allows the data aggregation to be represented in any desired areal unit. If
census data are reorganized this way and estimated to the areas where people likely reside, the
data will be less vulnerable to MAUP during analysis.
Surface representation also allows for a graphic unit to be uniform across a region,
offering a more accurate cartographic representation. Menni’s (2003) main focus was to describe
a methodology for generating a surface-based representation of demographic data using the
dasymetric mapping technique.
18
As previously mentioned, dasymetric mapping is a type of areal interpolation, it uses
ancillary data, such as parcel data, to aid in the areal interpolation process. In dasymetric
mapping the boundaries of cartographic representation are not random and they reflect the spatial
distribution of the phenomena being mapped. Menni’s (2003) used remote sensing ancillary data
to redistribute population to a raster grid. The example used by Menni’s (2003) covered five
counties in Pennsylvania, where there was a significant difference in the size of rural and urban
areas. Dasymetric mapping can be used to generate a surface model that is a more accurate
representation of the population in rural and urban blocks.
Walkability studies usually rely on many factors, but as mentioned above not all of them
produce reliable results. Dasymetric mapping was also an integral of this study. Like Menni’s
(2003) and Parsons (2015), dasymetric mapping is the considered most accurate procedure for
calculating parcel per population decreasing errors from that cause MAUP. This way the
distributions of the population per parcel can be close to the spatial phenomena.
2.4 Summary
To better understand the relationship between parcels and accessibility, of the schools under
consideration in this study, there must be an analysis that measures accessibility to schools. The
techniques utilized in this thesis employed some of the concepts explained in this chapter, and
goes further by utilizing roadways classification to derive accessibility.
It is the main objective of this study to research the extent of accessibility to schools that
avoids the problems inherent a buffer analysis and takes into account environmental and urban
barriers. For the size of San Juan County datasets, GIS tools were appropriate for this type of
analysis. As explained in the next Chapter 3 Methodology, the technique proposed in this study,
takes into account the size of the area studied and number of school sites to be represented.
19
CHAPTER 3: METHODOLOGY
GIS techniques to assess the effect of school closures on pedestrian accessibility are discussed in
this chapter. The data needed to carry out the analysis and present the method to combine parcel
and census data to dasymetrically map student population on the closed school inside the San
Juan County. Measuring the relationship between physical location of schools and parcels was
accomplished by utilizing some of the basic tools in ArcMap. The first subsection of this chapter
discusses the overall framework of the study, presents variables needed, and defines formulas
used to calculate the desire output. Some of these workflows will be presented as they were
constructed in Arc GIS ModelBuilder to accelerate the processing of some of the data and the
overall analysis of the studied area.
The analysis used in this methodology tests how closing of elementary schools in San
Juan County, Puerto Rico affected pedestrian accessibility for student population. Network
analysis procedures were used to quantify specified walking distances, defined as service areas,
between schools and residential parcels. The number of children within contrasting service areas
was compared before and after school closures. Furthermore, an origin/destination (OD) analysis
was performed to measure average distance walked to school for each child in the district in
order to assess which are the most accessible schools in the San Juan school district. It is
common for parcel data to have residential property information within its attributes. This was an
important and unique part of methodology developed for this study, as described below.
Walkability areas around schools were calculated using service area analysis tool from
ArcMap. To create the service area, a network dataset was built using accessible roads. Then
populations per parcel and service area were combined to determine the affected student
population before and after school closure. Finally, to assess the OD distance between schools
20
and parcels the OD Matrix Cost tool was used. This tool determines the shortest distance along
the network from the origin to the destination feature.
This executed methodology generated a walkability, and cadastral approach to the closure
of the schools in San Juan County. To develop this analysis, the following steps were completed:
1) extracted single and multifamily parcels; 2) assigned student population per parcel using
census block data; 3) identified pedestrian accessible roadways to create a network dataset; 4)
select parcel inside walkability area to assess school closure; 5) completed origin destination
analysis to assess mean distance to school; and 6) generated map visualization for review and
assessment. Figure 2 provides a detailed outline of this workflow.
21
Figure 2 Overview diagram of study workflow
3.1. Data Selection and Output Products
The input data utilized in this study is summarized in Table 1. The output of this study is
dependent on the quality of the information used to calculate student population per parcel and to
create the service area that represents walkability to and from schools (Table 1). Various maps
22
were generated to calculate student population per parcel and walkability in schools combining
all outputs of data. The datasets used in this study were drawn from Puerto Rico’s Sistemas de
Datos Geográficos (Puerto Rico Geographic Data Systems). These data included parcels, US
census blocks, roads, and schools.
The population census blocks contain information on the demographics in Puerto Rico
(Table 1). This data was used to calculate population per parcel. School data was also obtained
from Puerto Rico’s Sistemas de Datos Geográfícos. This data set contained all schools in point
geometry for Puerto Rico. Information on each school contained the classification of the schools
as stated by DE (Elementary, Middle, and High school). This data was used as the facility center
for the service area that derive walkability polygons.
Parcel data was obtained from El Centro de Recaudación de Ingresos Municipales
(CRIM 2015). This organization is in charge of tax and land information in Puerto Rico. This
parcel data contains single and multifamily parcels in Puerto Rico. This data was used with the
census block groups to dasymetrically map the study area. Thus dasymetric mapping was used in
areal interpolation including ancillary data such as parcels and census block data, which aided in
the calculation of student population per parcel.
Parcel data is an integral part of this analysis (Table 1). Thanks to dasymetric mapping it
is possible to add student population to each residential parcel in San Juan School district. For
this study dasymetric mapping was designed to calculated population below the age of 18 years.
The steps necessary for completing this analysis are outlined in the following sections. By
assigning student population to each parcel, it was possible to identify the areas where student
populations were most affected by school closure.
23
The roadways dataset, tiger roads, was obtained from the Sistemas de Datos
Geográficos (Puerto Rico Geographic Data Systems) and was the primary input for the networks
analysis that generated the service area (Table 1). Roadways were classified in order to identify
accessible roadways for pedestrians based on FHWA classification. Inside the attributes of the
tiger roads dataset, the road type classification was identified as stated by the FHWA.
ArcMap provided most of the necessary tools to organize key data layers for this study
for the analysis. The primary tools used in this study included “select by location,” “select by
attributes,” “spatial join,” and “summarize statistic.” The select by location tool allowed parcels
inside the walkability area to be selected. The select by attributes permitted selection of features
based on their attributes. For this study, this tool was used to select single and multifamily
residential parcels. The spatial join tool joins brings together attributes of different feature, based
on their spatial relationship. This was used to merge census block data population to the parcel
dataset and polygons. Finally, the summarize statistic tool was used to gather student population
from attributes of the spatial join. The output of this processing provided the basis for a better
understanding of the effects of closing schools (Table 1).
24
Table 1 Source Data Output
Feature Class Source Output Data
Parcelas (2014) Puerto Rico CRIM (CRIM 2015) Household Units
CensusBlockGroup
(CensusBlockGroup
2013)
Puerto Rico Sistemas de Datos
Geográfícos (Portal Datos Geograficos
Gubernamentales, 2003)
Population
DOTACIONES_EDU
CACION_ESCUELA
S_PUBLICAS
(Schools 2014)
Puerto Rico Sistemas de Datos
Geográficos (Portal Datos Geograficos
Gubernamentales, 2003)
Schools
Tiger Roadways
Tiger_rds_2006se
Puerto Rico Sistemas de Datos
Geográficos (Portal Datos Geograficos
Gubernamentales, 2003)
Accessible
Roadways
3.2 Population per Parcel
To assess student populations per parcel, a dasymetric mapping technique was implemented
using census block and parcel datasets. This analysis depended on parcel dataset attributes,
parcel classification and census block population attribute, population under 18 years. The
following subsections provide an overview of how these calculations were derived.
25
3.2.1 Population per Parcel Calculation
Total population per parcel was calculated by combining cadastral parcel data with US census
data at the block level. Thesis method proposed by Maantayet al. (2007) as Cadastral-Based
Expert Dasymetric Systems (CEDS) and also used by Holt (2004) included areal interpolation, to
determine population in urban areas utilizing cell classification from raster data. Utilizing land
use data and raster image, population per cell was calculated. Parsons (2015) further developed
his analysis for measuring population density using parcel data for pedestrian access to parks.
Based on their research, total population per parcel can be calculated according to the equations
below.
𝐻𝑢 1 = 𝑃 1 × 1
𝐻𝑢 2 = 𝑃 2 × 𝑅
𝐻𝑢𝑡 = 𝐻𝑢 1 + 𝐻𝑢 2
In these equations P1= The number of single family parcels in a block unit, P2= The
number of multi Family parcels in a block unit, Hu1= Household Units Single Family, Hu2 =
Household Units Multi Family, R = estimated mean number of households on multi-family
parcel, and Hut = Total household units. For this study, R was estimated as 2 households per
multi-family parcel. Once total household units were calculated for each parcel, another set of
formulas was used to distribute the census block population of those person aged 18 and younger
across parcels in a given census block.
𝐻𝑢𝑝 = 𝐵𝑝 /𝐻𝑢𝑡
𝑃𝑝 1 = 1 × 𝐻𝑢𝑝
𝑃𝑝 2 = 𝑅 × 𝐻𝑢𝑝
26
In these equations, Hup = People per Household, Pp1= Population per Parcel Single
Family, Pp2= Population per Parcel Multi-Family, and Bp = Total Population 18 and Younger
per census block.
Single or Multi Family parcels included in this analysis were defined using the parcel
classification contained in the attributes of this dataset. The data classifies single-family parcels
as an individual freestanding, unattached dwelling unit, normally built on a lot larger than the
structure itself. Multifamily includes many categories, from side to side townhomes to high-rise
buildings (Meyer 2004). Once the classification for parcels was determined, the desired parcels
were isolated and joined with the census block data (Figure 3).
Figure 3 ModelBuilder Showing Selection of Parcel and Spatial Join
3.2.2 Dasymetric Mapping
Dasymetric mapping is a technique to map data that are not equally distributed in order to
minimize the distribution mistakes presented in MAUP. With the use of ancillary data, parcel
data in this case, population per parcel was calculated. Utilizing two datasets, parcels and census
block group, student population per parcel was estimated using the spatial join tool in ArcMap
27
(Figure 4). Using this resulting attribute information, field named Pp1_Pp2, a closer look at eh
map can be seen in Figure 5. The darker shades symbolize the higher student population per
parcel while the lighter shades symbolize the lowest student population per parcel.
Figure 4 Student Population per Parcel Calculated
28
Figure 5 Student Population per Parcel Distribution
29
3.4 Walkability
The network analyst extension in ArcGIS functionality can be used to assess the quickest route
from one point to the other, and also service areas for facilities. For example, what stores are
within a certain distance when travelling by road from consumers’ homes? The use of network
analyst is beneficial to determine the least-cost network path between several origins and
destinations. The network path, derived from the road network layer consists of a system of
interconnected elements, such as lines and points, which represent the possible routes from one
location to another. For this report, a network dataset was created from accessible roads
identified within the Tiger roadways dataset. This data was the basis for generating the service
areas that establish walkability around the schools.
3.4.1 Accessible Roadways
Accessible roadways are defined as those that allow pedestrians to walk to and from locations.
All roadways have a classification that distinguishes them based on their characteristics. These
classifications are based on the meticulous criteria followed by any planning transportation
department (FHWA 1972). Because Puerto Rico is part of the US, the Puerto Rico Department
of Transportation utilizes US roadway classification standards (Departamento de Transportación
y Obras Públicas, 2015).
For this research, accessible roads were selected from Puerto Rico roadways dataset
following the FHWA functional classification. The FHWA is the organization responsible for the
construction, maintenance, and preservation of the Nation’s highways, bridges, and tunnels (U.S.
Department of Transportation 2013). This organization is responsible for describing the
functional criteria for road classification. This functional classification is based on a variety of
data points that indicate whether the roads are major network connectors or minor local roads.
30
These network connectors in road classification are influenced by mobility and access. Access
deals with means of entry, and exit on a road, while mobility deals with the congestion from
vehicles. Because most of the travel happens on a network of roadways, the functional
classification defines particular roadways segments within a hierarchy according to the
characteristic provided by the roadway. This classification was originally used for vehicles, but it
can also be used for pedestrian access. As it explained further below, some of the classification
for the roadways describes walkable pedestrian conditions.
The roadway hierarchy comes in three major categories (Figure 4): 1) Arterials are roads
that provide a low level of accessibility, 2) Collectors are roads that balance between mobility
and access, and 3) Locals are roads that provide a high level of accessibility. Based on this, the
functional classes are further divided as follow: 1) Interstate, 2) Other Freeways and
Expressways, 3) Other Principal Arterial, 4) Minor Arterial, 5) Major and Minor Collectors, and
6) Local Roads.
Interstates are the highest in the classification of arterials and are designed for large
mobility and longer distance travel. Other freeways and expressways look very similar to
interstates but are limited by physical barriers, access and exit points, and on- and off-ramps.
Other principal arterials serve as major centers for metropolitan areas. They have a high degree
of mobility but they also provide access to urban areas and driveways in specific parcels. Minor
arterials service moderate length trips and serve geographic areas smaller than its arterials
counterparts. Arterial roadways do not have sidewalk and walking on these roadways is not
encouraged and often illegal.
To create the network dataset that will support walkability, only roads accessible to
pedestrians were selected from the FHWA classification. These roadways were collectors and
31
local streets, which play a critical role in the network. Collectors are roadways that gather traffic
from local roads and funnel it to arterial roadways and are usually accessible for pedestrians.
Both urban and rural areas have collector roadways. Local roads are the largest proportion of
linear miles of all the roadways. They have short lengths, public transportation does not run on
them, and these roadways are often designed to discourage through traffic. Collectors and local
roadways usually have sidewalks and their speed limit is not high, thus are very accessible and
easy to walk along. Because of the accessibility of collectors and local roadways they comprised
the main input for the network data set used to create the service area for walkability (Figure 6).
32
Figure 6 Walkable Roadways in San Juan County
33
3.4.2 Service Area Methodology
Once the accessible roads were selected from the road dataset, a network dataset feature was
created. Using the network feature dataset created from the accessible roadways, a service area
was created using the network analyst tool in ArcMap (Figure 7). Service areas help evaluate
accessibility to and from a facility. The initial service area was created away from the school
within a distance of 0.25, 0.5, 1, and 2 miles from closing schools and recipient schools. This is
considered the average distance children usually walk to school (Horeck et al. 2012).
Figure 7 ModelBuilder Showing Selection of Parcel and Spatial Join
3.4.3 Walkability Output
The output of this study is a service area in the form of a polygon that shows the walkable areas
surrounding schools in San Juan County. This analysis was conducted on each school closed
(73), both before and after school closure (Figure 8 and 9). In combination with population per
34
parcel, the service area polygon was used to determine pedestrian accessible areas and
population affected by the closing of schools.
35
Figure 8 Walkability Area before School Closure
36
Figure 9 Walkability Areas after Closure
37
3.5 Student Population Affected
After the student population per parcel was derived and walkability area was calculated for each
school. The student population affected by school closure was determined. Using “Select by
Location” tool in ArcMap the parcels inside each section of the walkability area 0.25, 0.5, 1, and
2 miles were select and exported. Once completed using the “Summarize by Statistic” option, the
total student population affected was calculated in each section from the attributes in the dataset.
This method was repeated for each school in the San Juan County (Figure 10).
Figure 10 Parcels Affected Inside Service Area
38
3.6 Origin Destination Cost Matrix Analysis
The OD cost matrix analysis is a network analysis tool within ArcMap that calculates the least
cost distance for a road feature. The tool is used to compute a table that contains the total shortest
distance traveled from each origin to each destination.
The primary input for the walkability analysis was the network dataset containing the
accessible roads. Once the network dataset was processed the output is a series of line features
from origin to destination (Figure 11). The lines connect the origins to the destination but do not
represent the actual paths that were calculated, which give the shortest distance along the
accessible road network to the destination. This data is stored in the attributes of the dataset. For
example, Figure 11 shows the destination to facility lines produced by the OD tool.
Figure 11 OD Cost Matrix Analysis of a School Facility
39
The purpose of the OD analysis is to measure the mean distance walked to school before and
after school closure. All paths to schools in the county, in miles (Td), were added and divided by
the total student population for the respective school walkability area (Pw) to represent the
average journey (Aj) a student travels to school.
𝐴𝑗 = 𝑇𝑑 /𝑃𝑤
Using the calculated average journey values, histograms were created for each school before and
after school closure in order to assess the change in average distance walked. The primary
purpose of the histograms is to create visualizations of the data for results that can meaningfully
inform decision-makers in the school district about the consequences of the decisions to close
given schools on walkability. Further, the visualizations demonstrate the community impact of
the disappearance of "neighborhood schools." The results of the analyses described in this
chapter are presented in detail in the following Chapter 4 Results.
40
CHAPTER 4: RESULTS
The main goal of this study was to develop techniques to assess how the student populations in
San Juan County were affected by the closure of the elementary schools. The techniques
developed in Chapter 3 are required because a road network provides the spatial relationship
between parcels and elementary school locations. It is the job of the network analysis tools in
ArcMap to utilize this spatial relationship to derive a potential solution in the form of tables or
features that can be evaluated.
This chapter reviews the lessons learned using the Network Analyst extension in ArcMap
to create the service areas, in performing the Origin Destination Cost Matrix (OD) analysis.
Various sections in this chapter cover the comparison of the student population accessibility
derived from the walkability assessment of the schools, a narration of the schools that lost more
access and gained more access, and the mean percentage access within ¼ miles. A discussion
about landscape elements that might change the accessibility, and reports on a measure of the
changes in the average distance walked before and after school closures is also provided.
4.1 Network Analysis Issues and Solutions
To create the origin destination analysis, a network dataset was created from the accessible
roadways. When the tool ran for the first time a failure to return locations appeared. The error
was due to the Origin Destination (OD) cost matrix solver running out of memory to process a
10k by 10k matrix. To avoid this, the tools had to be run on each school individually.
For the service area analyses there was an output of 73 school walkability polygons
before school closure and 71 after school closure. Each polygon was made from four different
distances boundaries ¼, ½, 1, and 2 miles. In order to select the parcels with access to the school
in each location, the select by location tool was used to query the distances layer for each school.
41
For this study, although ¼ mile accessibility was emphasized, the other distances were also
calculated.
4.2 School Accessibility Results
Walkability analysis made from the accessible roads served as the first layer of the analysis to
identify accessibility of the student population. By using the select by location tool in ArcMap it
was possibly to observe all parcels inside each walkability areas. As previously mentioned, this
information collection method was performed before and after school closure. Before school
closure, 8,738 students had pedestrian accessibility at ¼ miles. After the school closure, the
students who could walk to the school at ¼ miles went down to 8,504 (Figure 12).
Using the numbers obtained from the output and indicated above, the decrease in
accessibility for student population was approximately 3%. This means that 3% of the total
student population of San Juan County that live within ¼ miles of a school that was closed lost
accessibility and are outside the normal range of walkability.
42
Figure 12 Graph of 1/4 Mile Access Results
By visually checking Figure 13 and Figure 14, one can observe that students living in
parcels affected by schools closures had to move in order to be able to walk to the next closest
school that had not been closed their walkable service area. In the maps below some schools
have no service areas nearby, because they have no single or multifamily parcels around them.
Even though these schools do not have a student population within a 2 mile service areas, these
schools were not picked to be closed by the local government. As illustrated in Figure 15, school
closing did not have major changes on accessibility with the exception of those five schools.
8736
8504
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
Before Closure After Closure
Student PoP
1/4 Mile Access to School
43
Figure 13 Student Population Accessibility before Closure
44
Figure 14 Student Population Accessibility after Closure
45
Figure 15 Schools in San Juan Most Affected School by Closure
From the original 73 schools analyzed, only 5 demonstrated major changes in
accessibility after the elementary school closure (Figure 15). A major change is defined as more
than the average of the student population for each school. Interestingly three of them, Gustavo
A. Becker, Jaime Rosario and Las Virtudes showed an increase of their accessibility at ¼ miles
this is because students of the schools that were closed will attend to these schools. While the
other two schools, Dr. Pedro G. Goyco, and Victor Pares Collazo demonstrated a marked
decrease in their student accessibility at ¼ mile. For these schools, the idea of neighborhood
schools is in some sense lost.
Of these five schools most affected by the closures Dr. Pedreo G. Goyci lost the most
accessibility. Before the school was closed, 250 students had access to the school at ¼ mile,
while after the school closed only 50 students had ¼ miles access to school.
46
The results of this study indicate that the school closures did not have a major effect on
the County overall regarding accessibility at a ¼ mile access. Figure 16 and 17 illustrate the
student population accessibility before and after school closure.
47
Figure 16 Student Population at a 1/4 Mile Access before School Closure
48
Figure 17 Student Population at a 1/4 Mile Access after School Closure
49
4.3 Environmental and Urban Obstacles
Environmental or urban obstacles affect walkability of an area directly. For Leslie et al. (2006)
these obstacles define the landscape of the urban growth, shaping the way cities develop, thus
affecting the way people move from place to place. Inspecting an aerial image of San Juan
County identifies one major obstacle that may affect accessibility; it is La Laguna San José
(Figure 18). This is a natural lagoon that connects to San Juan Bay, with an area of 3.85 square
meters (Recursos de Agua de Puerto Rico 2012).
Figure 18 Laguna San José and Laguna Los Corozos
50
To travel from one side of the laguna to the other there is only one bridge, el Teodoro
Moscoso. This bridge has 6 lanes, no walkable sidewalks and a toll of $3.25 each direction.
There is no way to walk across, as the bridge was not constructed for pedestrian use. As shown
in the Figure 18 above some of the school closed to the edge of the lakes have their service areas
interrupted, meaning that accessibility across the laguna to on those schools is prevented.
4.3. Changes in Average Distance
The results of this study indicate that overall walking distance to schools in San Juan County did
not change significantly due to the school closures. The OD Cost Matrix Analyst tool was used
to calculate the shortest distance from an origin, the parcel where student or students live, to an
origin location, and the schools they attend. The output of this analysis was a feature line that
provided the total distance walked to school before and after school closure (Figure 19). Before
the school closures the total distance walked to the 73 schools was 25,005 miles, and after
closure the total increase to 28,039 miles. This is an increase of 3,033 miles. Using this
information the increase was calculated for the mean distance walked to school.
51
Figure 19 Mean Distance Walked Before and after School Closure
This information was calculated by using the following formula:
𝑀𝑑 =
𝑇𝑑𝑥𝑃𝑝 𝑇𝑝
Md equals mean distance walked, Td is total distance walked for that school, Pp is the
student population of a parcel, and Tp is the total population of the service area for those schools.
Using a spatial join between the feature dataset from the output of the OD analysis and the
parcel, a new feature class was derived that displayed the distance walked before and after school
closure as shown in.
In Figures 20 and 21, one can observe at the parcel level an estimate of areas of the city
where students have particularly long or short walks to school. Also, by comparing the maps it is
possible to see that no particular area of the city suffered more of the burden of the school
closures on pedestrian accessibility. The relative walking distance to school of most areas of the
52
county to school remained the same. As displayed in these figures, there is a relative high total
distance walked to schools in the southern part of San Juan County. This is due to the fact that
the south side of the county us less urbanized than the north of the county with fewer accessible
roads to walk along to school.
53
Figure 20 Mean Distance Walked Before School Closure
54
Figure 21 Mean Distance Walked After School Closure
55
There were some individual schools that were affected by the school closure. One of
them was San Agustín elementary school. Figure 22 displays the histogram of the total distance
walked to schools before closure. The form of the histogram is symmetrical, meaning that the
mode, mean, and median for the distance walked to school is identical. In Figure 23, the
histograms is skewed. The distribution of the total distance walked to school reduces after school
closures. In this case, students actually cluster closer to the school after the closure of adjacent
schools.
Figure 22 Histogram of Total Distance Walked Before School Closure
Dataset 10
Frequency 10
0.02 0.67 1.33 1.98 2.63 3.28 3.93 4.59 5.24 5.89 6.54
0
0.35
0.69
1.04
1.38
1.73
Count
Min
Max
Mean
Std. Dev.
: 857
: 0.002177
: 0.65416
: 0.31504
: 0.11816
Skewness
Kurtosis
1-st Quartile
Median
3-rd Quartile
: 0.085658
: 2.9601
: 0.2372
: 0.31045
: 0.39754
Histogram
Transformation: None
Dataset : San Agustin Before Closure Attribute: Total_Length
56
Figure 23 Histogram of Total Distance Walked After School Closure
4.4 Scenario Analysis
The initial criteria designs by the Puerto Rico Department of Education (DE) for choosing which
schools to close did not include proximity to other schools. The proximity that two schools have
to each other could potentially change school accessibility at ¼ mile and the total distance
walked to school in San Juan County if a different set of schools were closed. To explore this
point, two schools that were close together were chosen for further analysis. These schools were
not included in the original wave of school closures. For this validation analysis, the original
schools that were closed were hypothetically modeled to remain open and the schools that
remained open were hypothetically modeled as closed (see Figure 24). This maintains the same
number of school closures overall. The originally closed schools were Luis Pereira Leal and
Manuel Cepero, as seen on the map in Figure 25. It is important to note however that theses
Dataset
Frequency 10
0 0.3 0.59 0.88 1.18 1.47 1.77 2.06 2.35 2.65 2.94
1.16
2.32
3.48
4.64
5.8
Count
Min
Max
Mean
Std. Dev.
: 2078
: 0.002177
: 2.9417
: 0.8693
: 0.6757
Skewness
Kurtosis
1-st Quartile
Median
3-rd Quartile
: 1.0305
: 2.89
: 0.34941
: 0.66796
: 1.2072
Histogram
Transformation: None
Dataset : San Agustin After Closure Attribute: Total_Length
57
schools were not as close together as Manuel A. Perez and Bolivar Pagan both of which were
analyzed as open.
58
Figure 24 Case Scenario of School Closure at ¼ mile access
59
The same analysis performed in the original study was performed on this scenario. For
the ¼ miles accessibility the total student population affected was of 8,736 before school closure
and 8,707 after school closure (Figure 25). The percentage of decrease for this scenario was less
0.3%. This indicates that closing schools that are in close proximity to one another does no
change the ¼ mile accessibility for the students.
Figure 25 Bar Graph 1/4 Mile Access to school before and after school closure scenario
The OD analysis was again performed in this new scenario before and after school
closure. In this scenario, the total distance walked to schools after school was the same as the
total distance walked after school closure (Figure 26). Unlike the original criteria, closing those
schools that were together did not result in a signficant increase or decrease on the total distance
walked to school before and after school closure.
8736 8707
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
Before After Scenario
Student PoP
1/4 Mile Access to School
60
Figure 26 Bar Graph Total Distance Walked to School Scenario
61
CHAPTER 5: DISCUSSION AND CONCLUSIONS
This study set out to develop a technique using the Network Analyst extension on ArcGIS, to
assess the effects of elementary school closures in San Juan School county, after the Department
of Education (DE) decided to close schools that fit their new criteria for closure in the “Plan de
Transformación y Reorganización de Escuelas” (Reorganization and Transformation School
Program, PRTE). The assessment performed in this study was completed using the Network
Analyst extension and Origin Destination (OD) Cost Matrix tools in ArcGIS Software. Network
Analyst was used to discern the accessibility of students to schools by creating a service area
around the location based on a network dataset, and the OD analysis tool was used to estimate
the shortest distance each student walked to school. This analysis was made before and after
school closure for a dataset of 73 schools, and the output was presented in the form of datasets,
tables and maps. The study shows the usefulness of Network Analysis and Origin Destination
(OD) Cost Matrix in the assessment of school site planning. This study also identified the
potential development of an automatic tool that could be used for school planners to model
school sites locations using the ModelBuilder extension in ArcGIS.
5.1 Benefits of Using Network Analyst and OD Cost Matrix
Most studies that use Network Analyst and OD analysis usually deal with park access as in
Parsons (2015) and Javed et al. (2013), but there is little information about the use of these tools
for school site planning. As of today, the general benefits of GIS for school site planning are
heavily explored (Cropper 2003), but no current studies utilize the methods presented in this
study to research schools accessibility, using network and population datasets.
It is also the first time a GIS study of this type has been performed for Puerto Rico.
62
In the past two years, the Puerto Rico DE used the following criteria to decide the school
closure: physical condition of the school, academic performance, utilization, and student
enrollment (Departament de Educacion de Puerto Rico 2014). And from all of these criteria,
only enrollment has an implicitly spatial component. Enrollment deals with the student
population that has access to school depending on how attendance boundaries or school
assignment zones are drawn. While enrollment was taken into account, the DE did not take into
account walkability in terms of how many students lost or gained pedestrian accessibility before
and after school closure. The methods utilized on this research focus on the spatial aspect that
affected student accessibility to schools. The information and results involved in this study
contained parcel, road, and population datasets and after analyzing the data, the study yielded
information about the affected areas after student closure.
5.1.1 Accessibility Measures
Accessibility is usually defined as the ease with which a location or service can be reached
(Nicholls 2001). This can be a measure of the spatial health of a city and can define the quality of
access in a spatial setting. Studies involving accessibility usually have a geometric approach
(Nicholls 2001). This geometric approach involves taking advantage of the spatial environment
using buildings, roadways, sidewalks and other elements to measure accessibility of locations.
This study utilized this geometric approach to assess accessibility using the roadways as the
primary input to create a network dataset. The roadways were classified using the FHWA
functional classification. This approach successfully facilitated the identification of those areas
that were more accessible for students to walk to and from school. Combined with the service
area analysis the areas affected by the school closure were successfully determined. No other
studies in Puerto Rico have taken the FHWA road classification into account for accessibility
63
purposes nor to create the service areas used in this study Thus, it was the purpose of this study
to present that option for future research.
5.1.2 Service Area Analysis
The Network Analyst extension uses the roadway data and creates a network dataset that
compromised of lines and nodes. This was the principal input that allowed Network Analyst tool
to create service areas around the schools. Services areas were displayed as ring shapes
articulated by connecting the outermost points along the road network at given network
distances. These areas represented the pedestrian accessibility around the schools. In the future,
service areas can become more precise by adding more feature layers. For examples, adding
features like stop signs and streetlights to the network could help to depict safer versus less safe
intersections for pedestrian crossing. If these were added as decision points in the network
analysis, it might be possible to increase the accuracy of the service area estimates.
Overall, the service area analysis in this study proved to be a valuable tool in the
assessment of school site planning, by identifying students that have access within walking
distance of the schools. Without this type of spatial distribution measurement, accessibility to
school cannot be judged with spatial precision. It is important to distinguish this from a buffer
method of analysis. A buffer method analysis do not represent the actual accessibility when used
in this type of assessment, thus are not considered a good way to measure student accessibility to
schools (Javed and Ahmad 2013). This is due to the fact that buffers lack an available network.
Buffer areas are a measurement of a direct route from the center of the buffer to the outside,
while service areas used the road network.
64
5.1.3 OD Analysis
The OD analysis is often used in transportation planning analysis (Yang 2002), and its benefits
are well documented when dealing with accessibility (Parsons 2015), but none of these studies
have dealt with school site planning. The OD analysis tool on this study was able to accurately
measure total distance walked to school. As demonstrated in Chapter 4, the output of this
analysis was a table that contained the distance walked to school from each parcel in which the
student reside to the closest school in their location. With this information the average distance
walked to school was calculated and compare before and after school closure. This analysis
played an important part in measuring what was the change after school closure in the distance
walked to school. In this analysis it was assumed that students would always take the shortest
path when traveling to and from school, and it accurately represented the spatial distance from
origin to destination. In this study it was demonstrated that the OD analysis is an effective
assessment tool to measure distances walked to and from school from any given location.
5.2 Limitations of the Project
Due to the availability and quality of the data there were some limitations in this study. GIS data
in Puerto Rico is very scarce, and those organizations that host most of the data do not share it
that easily. Roadways and population data can be found on the Puerto Rico Sistemas de Datos
Geografícos (Portal Datos Geograficos Gubernamentales, 2003) web site, but the road data set
had to be fixed before running the analysis using topology rules. These are rules that can be
added to spatial features with a spatial relationship to check for errors in their geometry (ESRI
2013). Population data available through the above website was in an unusual shape. The
original data was taken from the Census Bureau and merged with statistical analysis of the
population in Puerto Rico. For these reason there were only two attributes with the information
65
of the population. These two attributes were the population over 18 years old and the population
under 18 years old. For this study the populations under 18 years old was used as the student
population.
In this study, the parcel data was the only original data well formatted. This data
belonged to El Centro de Recaudación de Ingresos Municipales (CRIM 2015). Even though this
data was in good shape, there were some attribute definitions that were difficult in translation.
Some of the terms used for Parcel Type classification were not the common word used in
English. After contacting the San Juan city planners, these attributes were clarified. Even though
the data had limitations, after some minor editing it was in a good state for the research.
If the original school dataset had included information on school assignment zones, this
would have improved this research significantly. School assignment zones delimit sectors that go
with a specific school and are used to assigned student to their schools based on the student
proximity to the school like other counties in Puerto Rico, San Juan County is supposed to be
divided into spatially smaller districts. These districts are then divided into sectors with the
general expectation that students attend the school closest to where they reside. Upon further
research and conversation with personal from the DE, it was noted that there are no official
school assignment zone maps. The current protocol of the DE is that the student will attend the
closest school from their house unless the enrollment of that school is at maximum. Based on
this general rule, this study approximated the school assignment zone boundaries by using the
closest fitting service area to each school. Therefore, the results of this analysis present a best-
case scenario. If the precise changes in school assignment zones could have been mapped, the
actual effects of school closures on student walking distances to each school might actually be
worse than estimated here.
66
To calculate the student per parcel populations some assumptions were made. Two
parcel types were selected from the parcel type dataset, namely single family, and multifamily
parcels. To distribute the number of children per parcel, an assumption was made that there was
double the number of student population per parcel in multifamily compare to single family. This
assumption was made following the dasymetric mapping techniques used by Mennis (2003),
Holt (2004) and Parson (2015). It is possible that in some areas of San Juan County the relative
density of children in multi-family versus single-family parcels is higher or lower than this, but
unfortunately there was no data available on which to confirm this assumption.
5.3 Future Research
In June 2015, the governor of Puerto Rico, Honorable Alejandro García Padilla announced that
the island would not be able to pay back the public debt they owned of 72 billion dollars ( El
Caribe 2015). This crisis affected all branches of the government, and of course the DE. It was
announced later by the DE that they will close 100 more school this year and that the plan is to
close 600 more school by the end of 2016 ( El Caribe 2015). So far, the same criteria used before
will be used to close the schools next year and this will affect hundreds of students. This study
could potentially serve as a tool for the DE to decide which schools are more suitable to close
based on accessibility and distance walked to school. With additions to the network dataset like
sidewalks, tree canopy and crime data, this analysis could help to preserve the walkability to
school for the children of Puerto Rico.
Another area for future research would be to develop the analysis carried out in ArcGIS
ModelBuilder. These models developed in this study are automated workflows that string a
sequence of processes to produce an output. With some work, the methodology presented in the
current investigation can be automated such as by combining the models into an ArcGIS Add-In
67
or Extension which could be easily shared with other Esri software users. By automating this
process it would be possible for less advanced GIS users to quickly run the analyses conducted
as part of this thesis research. In this way, perhaps users would be able to perform analysis on
school districts. This could be of great benefit for schools planners and school districts.
5.4 Final Conclusion
Even though data quality and availability limited this research, the analysis performed is
considered successful. Being able to accurately assess schools accessibility and distance walked
to school is a valuable factor in school site planning (Carey 2011). The processes used in this
research could be potentially automated and integrated into a criterion that the DE could use for
future decision-making.
With better data quality this research could prove to be of significant use for school site
planning. As the first of this type of research in Puerto Rico, these results could increase
awareness of the potential benefits GIS for improved decision making for the growing
investment of data would signify better ways of measuring accessibility. It was the intention of
this study to add to the literature regarding school site planning through the use of development
of a methodology for measuring walkability. The approach used in this study can promote
awareness of the importance of spatial analysis.
68
References
Bliesner, Jim, Steve Bouton, and Barry Schults. 2015. Walk San Diego. Accessed February 13,
2015. http://walksandiego.org/download_file/view/108/96/.
Butterworth, Iain, and Edwards, Melissah. 2006. "Measuring the Walkability of Local
Communities Using Geographic Information Systems Data." The 7th International
Conference on Walking and Livable Communities. Melbourne Australia: School of
Health and Social Development 7: 1-14.
Carey, Kelley D,. 2011. School District Master Planning. New York: Rowman & Littlefield
Publishers, Inc.
Centers for Disease Control and Prevention. 2011. "Morbidity and Mortality Weekly Report."
School Health Guidelines to Promote Healthy Eating and Physical Activity. Accessed
June 5, 2015 http://www.cdc.gov/mmwr/preview/mmwrhtml/rr6005a1.htm.
CRIM. 2015. "Parcelario de Puerto Rico" Centro de Recaudación de Ingresos Municipales.
Accessed March 25, 2015. https://www.crimpr.net/crimdnn.
Cropper, Matt. 2003. "GIS In School Facility Planning." School Planning & Management 2: 56-
60.
Dellinger, A.M., y C.E. Staunton. 1999. "Barriers to Children Walking and Bicycling to School."
Morbidity and Mortality Weekly 51: 701-704.
Departament de Educacion de Puerto Rico. 2014. Plan de Transformación y Reorganización de
Escuelas.Accessed February 5, 2015. http://www.de.gobierno.pr/51-noticias/2251-plan-
de-transformacion-y-reorganizacion-de-escuelas.
69
El Caribe. 2015. Puerto Rico Cierra más de 150 escuelas por crisis económica. Accessed June
15, 2015. http://www.elcaribe.com.do/2015/05/14/puerto-rico-cierra-decenas-escuelas-
debido-crisis-economica.
EPA. 2013. "Travel and Environmental Implications of School Siting." Environmental
Protection Agency. Accessed March 10,2015.
http://www.epa.gov/smartgrowth/pdf/school_travel.pdf. Accessed:.
ESRI. 2013. ArcGIS Resources. Accessed June 15, 2015.
http://resources.arcgis.com/en/help/main/10.1/index.html#//0017000000nq000000.
FHWA. 2001 "Transportation Characteristics of School Children." Federal Highway
Administration. Accessed March 25, 2015. http://www.fhwa.dot.gov/ohim/1969/q.pdf.
Instituto de Estadísticas de Puerto Rico. 2014. "Publicaciones.". Accessed January, 15 2015.
http://www.estadisticas.gobierno.pr/iepr/LinkClick.aspx?fileticket=m8j9w8ftbhE%3D&t
abid=165.
Javed, M.A., y S.R. Ahmad. 2013. "Assessment of Neighborhood Parks Using GIS Techniques
in Sheikhupura City." Pakistan Journal of Science 2: 296-302.
Kunstler, James H. 1994. The Geography of Nowhere: The Rise and Decline of America's Man-
Made Landscape. New York: Touchstone.
Longley, Paul A., Michael, F. Goodchild, David J. Maguire, and David W. Rhind. 2011.
Geographic Informationn Systems & Science. New Jersey: John Wiley & Sons, Inc.
Mennis, Jeremy. 2003. "Generating surface models of population using dasymetric mapping."
The Professional Geographer 5:31-42.
Meyer, Nancy. 2004. GIS and Land Records The ArcGIS Parcel Data Model. California: ESRI
Press.
70
National Household Travel Society. 2013. NHTS Version 1.0 CD (Preliminary CD).
Washington, DC: Bureau of Transportation Statistics.
Nicholls, Sarah. 2001. "Measuring the Accessibility and Equity of Public Parks: A Case Study
Using GIS." Managing Leisure 6: 201-219.
Parsons, Jonathan Edward. 2015. "Mapping Uniformity of Park Access Using Cadastral Data
Within Network Analyst in Wake County, NC." MS thesis, University of Southern
California.
Portal Datos Geograficos Gubernamentales. 2003. Sistemas de Información Geografíca.
Accessed April 25, 2015. http://www2.pr.gov/agencias/gis/Pages/default.aspx.
Recursos de Agua de Puerto Rico. 2012. "Recursos de Agua de Puerto Rico." Lagunas del
Estuario de San Juan. Accessed August 1, 2015.
http://www.recursosaguapuertorico.com/Lagunas-San-Juan.html.
Rodrigue, Jean Paul. 2013. "The Geography of Transport Systems." Methods in Transport
Geography. Accessed July 20, 2015.
https://people.hofstra.edu/geotrans/eng/methods/ch2m1en.html.
Speck, Jeff. 2013 Walkable City. New York: North Point Press.
U.S. Department of Transportation. 2013. "Federal Highway Administration." Highway
Functional Classification Concepts, Criteria and Procedures. Accessed March 5, 2015.
http://www.fhwa.dot.gov/planning/processes/statewide/related/highway_functional_class
ifications/fcauab.pdf.
Yang, Zhongzhen. 2002. "GIS-Based Analysis of Railway's Origin/Destination Path-Selecting
Behavior." Computer-Aided Civil and Infrastructure Engineering 3: 221-226.
71
Yin, Jie, and Qingming: Qian, Yi Zhan. 2014. "The GIS-based Research of Measurement and
Planning on Walkability of Educational Public Services Facilities - a case study in
Wuhan." Advanced Materials Research 926: 4037-4040.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
The role of amenities in measuring park accessibility: a case study of Downey, California
PDF
Investigating bus route walkability: comparative case study in Orange County, California
PDF
Smart growth and walkability affect on vehicle use and ownership
PDF
Measuring seasonal variation in food access: a case study of Everett, Washington
PDF
Network accessibility and population change: historical analysis of transportation in Tennessee, 1830-2010
PDF
Integration of topographic and bathymetric digital elevation model using ArcGIS interpolation methods: a case study of the Klamath River Estuary
PDF
Mapping uniformity of park access using cadastral data within Network Analyst in Wake County, NC
PDF
Using pedestrian accessibility indicators to locate schools: a site suitability analysis in Greenville County, South Carolina
PDF
Risk analysis and assessment of non‐ductile concrete buildings in Los Angeles County using HAZUS‐MH
PDF
Visualizing historic space through the integration of geographic information science in secondary school curriculums: a comparison of static versus dynamic methods
PDF
Analyzing earthquake casualty risk at census block level: a case study in the Lexington Central Business District, Kentucky
PDF
Assessing the impact of a web-based GIS application to promote earthquake preparation on the University of Southern California University Park Campus
PDF
A contributory Web-based application for documenting historic resources: French colonial era art deco architecture in Hanoi
PDF
Beach morphodynamic change detection using LiDAR during El Niño periods in Southern California
PDF
Analysis of park accessibility in Redan, Georgia Web GIS application
PDF
Preparing for earthquakes in Dallas-Fort Worth: applying HAZUS and network analysis to assess shelter accessibility
PDF
Finding the green in greenspace: an examination of geospatial measures of greenspace for use in exposure studies
PDF
Access to active play parks for youth segments in Alexandria, Virginia
PDF
Precision agriculture and GIS: evaluating the use of yield maps combined with LiDAR data
PDF
Questioning the cause of calamity: using remotely sensed data to assess successive fire events
Asset Metadata
Creator
Rodríguez, Joel
(author)
Core Title
Walkability study for school accessibility: case study of the San Juan, Puerto Rico elementary schools
School
College of Letters, Arts and Sciences
Degree
Master of Science
Degree Program
Geographic Information Science and Technology
Publication Date
09/21/2015
Defense Date
09/02/2015
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
geographic information systems,GIS,OAI-PMH Harvest,origin destination,pedestrian access,Population,Puerto Rico,school assessment,school planning,service area,walkability
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Vos, Robert O. (
committee chair
), Lee, Su Jin (
committee member
), Swift, Jennifer N. (
committee member
)
Creator Email
joelr@usc.edu,rodriguezjoelr@yahoo.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-185533
Unique identifier
UC11275309
Identifier
etd-RodrguezJo-3941.pdf (filename),usctheses-c40-185533 (legacy record id)
Legacy Identifier
etd-RodrguezJo-3941.pdf
Dmrecord
185533
Document Type
Thesis
Format
application/pdf (imt)
Rights
Rodríguez, Joel; Rodriguez, Joel
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
geographic information systems
GIS
origin destination
pedestrian access
school assessment
school planning
service area
walkability