You are on page 1of 19

Transportation and Recreational Infrastructure

Development: Transport and Land Use Measures


for Public Health Comparisons
Michelle R. Oswald Beiler, M.ASCE 1; Greg Miller 2; and Madelyne Brown 3
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Abstract: Understanding how transportation and recreational development projects impact local communities, specifically with regard to
public health, is critical for continued investment. The purpose of this study is to analyze the impact of transportation and recreational
infrastructure projects, specifically parks, pools, and trails, on local communities with a preliminary connection to public health measures.
Two objectives are achieved: (1) develop a process for analyzing transportation and recreational infrastructure using project-specific and
network-specific metrics and (2) implement the process through a pilot study in central Pennsylvania. Projects funded under Pennsylvania’s
Department of Conservation and Natural Resources Community Conservation Partnership Program (PA DCNR C2P2) within four counties
(Lycoming, Northumberland, Snyder, and Union) are analyzed. The results are synthesized at the project as well as the minor civil division
(MCD) level in order to reflect the opportunity for health measure comparisons. Highlights include strong positive correlations between
population density and trail demand metrics as well as pool demand and walkable household density metrics. Metrics are then further
analyzed in relation to neighborhood obesity rates. DOI: 10.1061/(ASCE)UP.1943-5444.0000468. © 2018 American Society of Civil
Engineers.
Author keywords: Transportation management; Public health and safety; Recreation; Geographic information systems.

Introduction and Background can be explored in terms of health impacts. In order to do so, two
primary objectives are achieved: (1) develop a process for analyz-
With goals of improving mobility while reducing impacts on the ing transportation and recreation projects using project-specific and
environment, economy, and society, federal and state agencies network-specific metrics and (2) implement the process to a pilot
have been supporting the need for transportation and recreational study in central Pennsylvania.
infrastructure development. For example, for over 20 years, the To meet the research objectives, this study investigates projects
Pennsylvania Department of Conservation and Natural Resources funded through the PA DCNR C2P2 grants, which fall into the sub-
(PA DCNR) has been providing funding through the Community categories of Trails and Park Rehabilitation and Development in
Conservation Partnership Program (C2P2) to support sustainable Lycoming, Northumberland, Snyder, and Union counties in central
community development and revitalization projects across the Pennsylvania. In order to further connect the projects to public
state (PA DCNR 2016). Projects such as parks, pools, and recrea- health measures, the counties were selected based on their location
tional facilities, as well as trail rehabilitation and trail develop- within the Geisinger Health System region (organization that
ment, have been funded. Although infrastructure projects are provides primary health care services in central and northeastern
being implemented, the impact of these projects on local com- Pennsylvania). Using geographic information systems (GIS), the
munities and public health is relatively unknown. Understanding pilot study projects (PSPs) are evaluated based on transportation
how transportation and recreational infrastructure projects impact and land use network and project measures. A preliminary connec-
local communities, specifically with regard to public health and tion to publicly available health data (obesity rates) is provided in
well-being, is critical for continued funding (CDC 2010). this study as an example for how health care services like Geisinger
The primary purpose of this study is to systematically analyze can use the project and network measures with public health data.
the impact of transportation and recreational projects, specifically
parks, pools, and trails, on local communities. A preliminary con-
Transportation and Health
nection to public health measures, specifically neighborhood obe-
sity rates, is provided as an example of how recreational projects Walkability has been identified by multiple studies to improve
public health (Litman 2003; CDC 1999). Walkable communities
1
Associate Professor, Dept. of Civil and Environmental Engineering, have shorter blocks, pedestrian-oriented streetscapes, sidewalks,
Bucknell Univ., 1 Dent Dr., Lewisburg, PA 17837 (corresponding author). lighting, and crosswalks, and, overall, are more pedestrian friendly,
Email: michelle.beiler@bucknell.edu increasing the demand for walking (Adams et al. 2016). In particu-
2
Undergraduate Research Assistant, Bucknell Univ., 1 Dent Dr., lar, walkable neighborhoods have a 35% lower risk of obesity
Lewisburg, PA 17837. Email: glrm001@bucknell.edu and may reduce the risk of diabetes by more than 10% (Bakalar
3
Undergraduate Research Assistant, Bucknell Univ., 1 Dent Dr.,
2016). Samimi and Mohammadian (2010) found that for every
Lewisburg, PA 17837. Email: mmb021@bucknell.edu
Note. This manuscript was submitted on February 15, 2017; approved 1% decrease in automobile usage, the chance of obesity is re-
on April 6, 2018; published online on July 24, 2018. Discussion period duced by 0.4%. Braun et al. (2016) explored walkability in connec-
open until December 24, 2018; separate discussions must be submitted tion to health impacts such as cardiometabolic risk factors and
for individual papers. This paper is part of the Journal of Urban Planning found a relationship between increased walkability and low blood
and Development, © ASCE, ISSN 0733-9488. pressure rates. Schauder and Foley (2015) determined that active

© ASCE 04018030-1 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


transportation contributes to reductions in cholesterol levels, and surrounding communities because they often provide safe, open
Falconer et al. (2015) explored the association between active travel spaces for physical activity. A pool is a type of park in which there
patterns in school-age boys and lower body mass index. is a still body of water intended for swimming. Often, large park
Though active travel has been shown to have significant health developments also have pools. The facilities of a park must cater to
benefits, specifically 30–45 min of walking or biking (CDC 1999), the specific values of a community; thus, the efficacy of a park is
only about 4.3% of adults actively commute to work (CDC 2015). mainly determined by the potential for different choices and oppor-
Even with high-quality walking and biking infrastructure present, tunities for activity (Harnik and Welle 2011).
individuals may still choose not to participate in active travel, sug- Previous studies have shown that physical activity establish-
gesting additional efforts are needed. ments are not evenly distributed geographically, nor evenly dis-
The decision to commute by active travel is influenced by many persed among all people. Although research shows that people
factors, such as the proximity to such infrastructure (Heinen et al. who live within a half mile of a park tend to exercise more than
2015) and personal physical activity level (Meloni et al. 2009), as people who lack park access (Urban Land Institute 2015), only
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

well as the commute environment. For example, zoning and develop- 35% of adults live within a half mile of a park (CDC 2015). A study
ment regulations (Cannon et al. 2013), including the presence of by the Center for Rural Pennsylvania (2014) showed that 9% of
small retail and smaller lot sizes, have been shown to be positively Pennsylvanians live in a municipality without local parkland.
associated with the likeliness that an individual will walk or bike Therefore, exploring opportunities for development as well as the
(Appleyard 2012). Although more research is needed, connected impacts of existing development is valuable.
and efficient infrastructure is at the core of promoting active travel
(Buehler et al. 2016).
Recent tools and methods have been developed to quantify the Physical Activity Facility Demand
benefits associated with active transportation. The Health Impact The overall impact of a physical activity development is directly
Assessment (HIA) is a tool that analyzes the potential impacts related to the facility demand. A 2009 study demonstrated how
of projects ranging from physical activity to mental health to in- the implementation of a high-quality multipurpose trail failed to
form transportation planners of potential effects of their decisions draw in users from its surrounding neighborhood and was unsuc-
(Dannenberg et al. 2014). Similarly, the Health and Economic cessful in increasing physical activity (Burbidge and Goulias
Assessment Tool (HEAT) for cycling and walking is a tool that can 2009). Similarly, in the pursuit of increasing park demand across
be applied to plan new active transport infrastructure and provide demographics via renovations and rehabilitations, improvements to
a prospective health impact assessment (WHO 2014). Also, the parks may not automatically result in increased use and physical
US Department of Transportation published a Transportation and activity (Cohen et al. 2009).
Health Tool, which compares states and metropolitan areas to User demand for sustainable infrastructure is key to understand-
one another based on policies and various transportation statistics ing the magnitude of the health impacts. For trail demand, Oswald
(USDOT 2016). This tool provides statistics about the percentage Beiler et al. (2015) developed a method for assessing the impacts
of people who commute by bicycle and the use of federal funds for of rail trails, providing evidence for the use of infrared trail counters
bicycle and pedestrian efforts. to measure nonmotorized (primarily pedestrian and bicycle) travel
demand. The Federal Highway Administration (FHWA 2011) pro-
vides information on choosing a pedestrian or bike counting tech-
Physical Activity Facility Development
nology that focuses on the type of user, length of the study, and
Recreational facilities, including parks, pools, and trails, contribute cost of technology. The major difference between nonmotorized
to the physical activity spaces of a local community. Efforts to volume and vehicle counting is that nonmotorized volumes are
develop these facilities continue with state and local agencies pro- much more dependent and sensitive to environmental conditions,
viding funding incentives. For example, PA DCNR’s goal is to have causing counts to significantly vary by the hour, day, and time of
a trail within 15 min of every citizen in PA (PA DCNR 2017). The year (FHWA 2011). Therefore, automatic counting methods are
planning and development of safe and accessible local parks is the useful in gathering comprehensive data over time. Some of the most
overall priority of the Community Parks and Recreation Grants pro- common methods are passive infrared, active infrared, pneumatic
gram, and therefore, over the last decade, many local parks and tubes, inductive loops, piezoelectric sensors, and radio beams.
pools have been developed and rehabilitated (PA DCNR 2015). Passive infrared equipment, which was selected and used in this
This program encourages projects that advance local conservation study, detects passing people by comparing the temperature of
and recreational goals of communities throughout the state. the background to the heat emitted by the individual, whereas active
Previous studies such as Schwartz et al. (2011) have begun to infrared uses an infrared beam and receiver.
explore physical activity establishments in connection to health mea- In order to validate the automatic results, manual counts are rec-
sures, such as body mass index. For the investigation, Schwartz et al. ommended, which can be useful in determining adjustment factors
(2011) used eight categories for physical activity establishments: (1) as well as directional and mode splits (Ryus et al. 2014). In terms
fitness and recreational sports centers or sports and recreation of adjustments, when two people pass an infrared counter at the
instruction; (2) other amusement and recreation industries; (3) high same time, only one count may be recorded; therefore, comparing
schools; (4) boat access points and trailheads; (5) DCNR park proj- manual counts to automatic counts can determine the adjustment
ects; (6) preserved areas; (7) DCNR trails; and (8) miscellaneous factor. Also, expansion factors (seasonal adjustment factors), such
sites, including bowling, golf courses, country clubs, nature parks, as those developed by the National Bicycle and Pedestrian Docu-
and the Young Men’s Christian Associations (YMCAs). The mentation Project (NBPDP 2009), can be used to estimate usage
Schwartz et al. (2011) study serves as a foundation for this study over a longer time period.
because it establishes the value of exploring land use metrics in re- Similarly, a methodology for measuring the user demand of
lation to electronic health records, specifically body mass index parks is crucial in understanding the influence of the park on the
(BMI). surrounding community. Although advanced counting technolo-
Parks are open or enclosed public spaces developed for rec- gies for parks exist, they are expensive and overall challenging to
reational use. They play a vital role in the public health of their implement at the pilot study level. Other nontechnology methods

© ASCE 04018030-2 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


used to estimate park usage include manual counts and interviews GIS in which the user can string together data sets, geoprocessing
(Silver et al. 2014). tools, and other operations, such as an iterative tool that allows for
the repetition of a task for multiple features (ESRI 2016c). When a
model is complete, it is run, and the desired output, such as a table
GIS Metrics or a feature layer, of the model’s process is produced.
The use of geographic information systems in transportation plan- Though GIS is an extremely helpful tool in evaluating many
ning and network analysis applications can be extremely valuable. performance metrics, there are limitations and challenges when
GIS allows for spatial investigation of multiple networks and data exploring real-world problems. Data consistency, specifically with
layers through visualization. Similar to motorized transportation regard to infrastructure and land use in rural areas, was explored
performance measures, nonmotorized metrics can be used to mea- by Stewart et al. (2016). Finding ways to coordinate data sources
sure the effectiveness of a system. Semler et al. (2016) presents a across local jurisdictions is fundamental in ensuring accuracy. Also,
guidebook for developing pedestrian and bicycle performance met- the modified areal unit problem (MAUP) is “a problem arising from
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

rics that suggests that the metrics stem from community goals such the imposition of artificial units of spatial reporting on continuous
as connectivity, health, and safety and focus on the transportation geographical phenomena resulting in the generation of artificial
measures of accessibility, compliance, demand, reliability, mobil- spatial patterns” (Ervin 2016). In the process of grouping data
ity, and infrastructure. in GIS, information can be lost and can inevitably cause a bias
In order to quantitatively measure transportation networks, GIS (Ervin 2016). Dalton et al. (2014) explored the route differences
spatial analysis tools can be useful. Tresidder (2005) defines seven between those predicted by GIS and those actually taken by com-
connectivity metrics using GIS, including link-node ratio (number muters. Actual routes are typically 27% longer than those modeled
of links divided by the number of nodes), connected node ratio in GIS because people stop at intermediate locations and avoid
(number of real nodes divided by the total number of nodes), aver- crowded roads. However, the walking routes are 13% shorter than
age block length (sum of road length per unit area divided by the the GIS routes because people take paths that are not included in
number of nodes per unit area), and intersection density (number of the GIS network data set (Dalton et al. 2014). Zhang et al. (2011)
intersections per unit area). Similarly, Dill (2004) presents mea- analyzed the accuracy of the assumption that people will use the
sures of connectivity, including pedestrian route directness (PRD), closest facility. Although the study finds that this assumption is
which is computed by dividing the network distance from origin to not always true, Zhang et al. (2011) pointed out that it is an espe-
destination by the geodetic distance. PRD is described as the best cially convenient assumption to make when using GIS. These fac-
connectivity indicator because it reflects minimizing trip distances tors and limitations are addressed in this study because GIS is the
and optimizing route directness; however, it is challenging to com- primary tool used to analyze the metrics.
pute using GIS.
Many GIS studies have been successful in demonstrating the
effectiveness of using this technology to evaluate walkability and Methodology
connectivity. Forsyth and Oakes (2014) found that density mea-
sures and connectivity measures are positively correlated with The following six-step methodology was developed in order to
walking. Gori et al. (2014) presented various ways to measure the analyze the pilot study project based on the four-county region
quality, connectivity, and proximity of a road network and sug- in central PA (Lycoming, Northumberland, Snyder, and Union):
gested no single measure can accurately describe the walkability (1) project scoping, (2) project inventory, (3) metric identification,
of an area. Schlossberg and Brehm (2009) further provided GIS (4) GIS data collection, (5) field data collection, and (6) metric ap-
techniques in evaluating walkable environments, such as creating plication. The four-county region was selected as an example of a
a walkable street layer and analyzing pedestrian catchment areas subset of the Geisinger Health System service area in order to ex-
using a half-mile network and geodetic buffer, which is the theo- plore the method and determine applicability to a real-world net-
retical walkable zone around a point. work. Each step is discussed subsequently in terms of how it is
applied to the four-county region. Although this research focused
on those counties, the process was developed in order to be applied
GIS Tools at a broad level, inclusive of not only other counties within the
The use of GIS can greatly simplify the computation of common Geisinger service area, but counties throughout the nation.
metrics. The Network Analyst tool can solve a variety of network
problems, such as finding the closest facility, routing origins to des-
Project Scoping (Step 1)
tinations, and identifying service areas around a location, when
combined with a road network data set (ESRI 2016d). Many of In order to apply the method to a real-world application and sim-
the GIS desktop program capabilities, as well as new tools for ulate the process for comparing project development patterns with
streamlining computations, appear on ArcGIS Online. In ArcGIS health measures, the PA DCNR C2P2 projects, specifically those
Online, network analysis tools can be used with various predefined funded under the two grant categories, trails and park rehabilitation
travel modes such as driving, rural driving, and walking distances, and development, were provided by PA DCNR (2015). A database
all of which use the road network data set. Driving distances model of projects as well as a GIS shapefile of the project locations were
the movement of cars, with rural driving distance allowing cars to provided.
drive on unpaved roads, whereas walking distance follows only Because the research focus is a pilot study of select counties
roads that allow for pedestrian traffic (ESRI 2016b). Other GIS within the Geisinger Health System region, the projects that were
tools are available on the online platform, such as the Business An- identified as part of the research scope were those included within
alyst Web Application, which applies GIS to business data (ESRI the four-county region (Lycoming, Northumberland, Snyder, and
2016a). Union). Based on data availability, projects funded through the
When performing metric calculations and using large sets of C2P2 program between the years 2002 and 2015 were selected
data, the use of the GIS ModelBuilder tool can be extremely help- for the pilot study, reflective of Geisinger EHR data (available from
ful. ModelBuilder is a visual programming language specific to 2002 to present).

© ASCE 04018030-3 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Table 1. PA DCNR project inventory process
Number of projects (%) Filtering process
122 (100%) All projects outside the four-county region eliminated (original data set: 3,389 project line items)
−4 (3.28%) Four trails discovered for primary ATV/snowmobile usage
−3 (2.46%) Three identical duplicate points found in data set
−20 (16.39%) Nineteen projects fall outside of 2002–2015 date range; one project with missing funding date information
þ1 (0.0082%) One project had both park and pool improvements listed in the scope of renovation. This project was split into two projects:
park and pool renovations.
−2 (0.016%) Two more projects were at sites that have closed down
−17 (13.93%) Twenty-eight projects were listed as single phases of development/rehabilitation to facilities. Multiple phases of these
28 projects were condensed into one project phase per facility, leaving 11 individual projects and eliminating 17 projects
77 (∼63%) Approximately 63% of the original 122 within the 4 county region were kept in the final PSP analysis
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Note: Bold text indicates the total projects that remain after filtering.

Project Inventory (Step 2) and project-specific aspects such as demand. In order to group
PA DCNR (2015) provided 3,389 data entries for parks, pools, and the metrics based on similar goals, five categories were estab-
trails in Pennsylvania, all of which received some form of funding lished: (1) density, (2) connectivity, (3) physical activity facilities,
from DCNR through the C2P2 program. Of these 3,389 projects, (4) project demand, and (5) project magnitude. Table 2 includes a
122 of the geocoded points fell within the study’s four-county list of the transportation and land use metrics identified and devel-
region. Although 122 projects were identified in this four-county oped for this study. The metrics represent a combination of standard
region, a data filtering process (removal of project duplicates, non- geographical network measures (such as population density, inter-
applicable projects, and projects outside of the analysis timeframe) section density, and link-node ratio), as well as metrics that char-
was applied, as shown in Table 1. acterize the projects (such as trail demand, trail length, and pool
First, it was found that four projects were based on motorized facility area). By establishing metrics at both levels, there is the
trails for all-terrain vehicles (ATVs), rather than nonmotorized opportunity to explore correlations between the recreational proj-
(pedestrian and cycling) facilities; thus, they were eliminated be- ects and network accessibility. Each metric is further defined along
cause these trails were not within the research scope of facilities with the application level at the project, project buffer, or minor
supporting physical activity. Next, 3 projects were duplicated within civil division (MCD) level.
the data set, 19 projects had funding expiration dates either before The project level refers to applying the metric directly to the indi-
2002 or after 2015, and 1 project had no date information; therefore, vidual project location. The project buffer level refers to analyzing
all of these projects were removed. Next, it was found that two addi- the metric within a five-mile network buffer around the project ac-
tional projects were carried out at sites that had since closed, and cess points, unless otherwise specified. A bikeshed is defined by the
were consequently removed. There was one project that had both Federal Transit Authority (FTA 2011) as a 4.83 km (3 mi) buffer
park and pool improvements listed in the scope of the renovation. around a facility. Another study by Iseki and Tingstrom (2013)
This project was divided into two parts, the park and pool renova- on bikesheds defines a 7.08 km (4.4 mi) network buffer bikeshed
tions, thus adding one project to the data set. Finally, 28 projects based on average bicycle trip distance data and energy expenditure.
were listed as single phases of development or rehabilitation to fa- Therefore, for this research project, a network buffer of 8.05 km
cilities. Multiple phases of these 28 projects were condensed into one (5 mi) is used for a majority of the project measures to encompass
project phase per facility, leaving 11 individual projects. This came both the 4.83 km (3 mi) and 7.08 km (4.4 mi) in order to reflect
to a total of 77 projects, also referred to as the pilot study projects. bicycle accessibility, as well as to be inclusive to other modes of
As outlined in Table 1, the scope of the projects included in the transportation. Last, the MCD level refers to analyzing the metric
analysis was redefined based on the following assumptions about based on the minor civil division boundary (PASDA 2014b). For
projects included in the PSP analysis: example, the roadway network distance measure (average network
• Funded by DCNR between 2002 and 2015 (based on expira- distance from households to nearest project) is in the connectivity
tion date). category and is applied at the project buffer and MCD level because
• Complete funding date information available. it is based on the network, whereas the trail mode split (ratio of
• Not identified as ATVor snowmobile (motorized) trails or parks. pedestrians to cyclists) measure is under the project demand cat-
• Open facilities that are not closed or planned for closure for egory and is only applied to the project itself.
2015 and 2016. The only metric that was applied at a smaller project buffer
• Expiration dates within 2002–2015 (inclusive)—It was assumed level, not based on 8.05 km (5 mi) buffer, was walkable household
that the “expiration date” for the project was also the “opening
density. Walksheds, which are well-established compared to bike-
date” for a project in order to ensure that the entire facility was
sheds, are traditionally defined as a quarter-mile to half-mile buffer
fully functioning and available for use at that time. Although
around a facility (Zhang et al. 2011). Pedestrian catchment areas
some parks may have had an opening date before the reported
and walkable zones tend to be half-mile buffers around a point;
expiration date, there was no information to ensure that the fa-
thus, in order to reflect the traditional walkshed area, a half-mile
cility was complete or officially publicly accessible before then.
buffer around the project is used for the walkable household density
For projects with multiple phases, the expiration date for the first
phase was used. metric (Schlossberg and Brehm 2009).

Metric Identification (Step 3) GIS Data Collection (Step 4)


Based on the goals of the study, transportation and land use metrics The next section discusses the use of geographic information sys-
can be used to analyze characteristics such as density, connectivity, tems to gather spatial data in preparation for applying the metrics.

© ASCE 04018030-4 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Table 2. Metric definitions and application
Category Measure Application Definition
Density Walkable household Project buffer ½ mi Number of households within a network distance of ½ mi
density from a project (project buffer based on ½ mi)
Household density Project buffer, MCD Number of households per unit of area
Population density Project buffer, MCD Number of people per unit of area
Connectivity Roadway network Project buffer, MCD Average network distance from households to nearest project
distance
Geodetic distance Project buffer, MCD Average linear ground distance from households to nearest project
Route directness Project buffer, MCD Roadway network distance divided by the geodetic distance
Link-node ratio Project buffer, MCD Number of links divided by the number of intersections (real nodes)
and cul-de-sacs (dangle nodes)
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Average block length Project buffer, MCD Number of street intersections divided by the number of intersections
plus cul-de-sacs
Intersection density Project buffer, MCD Number of intersections per unit of area
Connected node ratio Project buffer, MCD Number of street intersections divided by the number of intersections
plus cul-de-sacs
Roadway density Project buffer, MCD Number of linear miles of street per square mile
Physical activity Physical activity Project buffer, MCD The ratio of the number of types of physical activity facilities present in an area.
facility facility Physical activity facilities include fitness and recreational centers (ESRI Business
development diversity index Analyst NAICS code 713940), other amusement and recreational industries
(ESRI Business Analyst NAICS code 713990), schools (ESRI Business Analyst),
boat and trail access points (PASDA: DCNR Explore PA trails—Trails (points)
PASDA: PFBC Access Points (Fishing and Boating)), DCNR park projects and
local parks (DCNR Open Data: Local_Parks_2014), preserved areas (PASDA:
Protected Land Inventory), DCNR trails (DCNR Open Data:
Pennsylvania_DCNR_Bureau_of_Recreation_and_Conservation_Grants_Trails),
and miscellaneous: bowling centers, golf courses and country clubs, nature parks,
other similar institutions, YMCAs (ESRI Business Analyst NAICS codes 713910,
712190, 813319, 713950) Searched by title “YMCA” in Business Analyst).
Non-PSP density Project buffer, MCD Number of non-PSP facilities per unit area
PSP density Project buffer, MCD Number of PSP per unit of area
Project demand Trail mode split Project Ratio of pedestrians to cyclists
Trail demand Project Number of trips per year
Pool demand Project Number of attendees per season
Park demand Project Number of attendees per year
Project magnitude Trail length Project Length of the trail in miles
Pool facility area Project Acreage of the pool facility
Park facility area Project Acreage of the park facility
Note: Project Buffer = project buffer of a 5-mi network distance around the project (except for walkable household density, which is based on a ½ mi distance);
MCD = minor civil division; and Project = application to the individual project.

Multiple data sets were collected and are described based on the of an access point layer, later to be used in Network Analyst. The
following categories: project data, county data, and network data. vast majority of these data could be found in the Local Parks layer
within the online open resource DCNR Open Access (PA DCNR
Project Data 2014). Any missing data had to be manually digitized through fur-
The spatial data provided by PA DCNR for the projects included ther site research. This was carried out through site identification
location; however, it was based on a specific point rather than the through the DCNR project point data, followed (as thoroughly as
actual project delineation (such as a line for a trail or polygon for a possible) by site boundary examination of Google Earth (2016) and
pool or park). The acquisition of trail data in the form of GIS digi- Google Maps Street View (2016).
tized line representation was essential for network analysis. Of the Once polygon project data were collected or digitized, a similar
seven trails examined in the pilot study, trail data were found for inventory of access points for each project was completed. The
only two (Buffalo Valley Rail Trail and Susquehanna Riverwalk) identification of access points was helpful to create network buff-
within open source resources such as DCNR Open Access and the ers, service areas tending to the recreational demands of popula-
Pennsylvania Spatial Data Access (PASDA 2013). Thus, the re- tions surrounding each project. An examination of Google Earth
maining five had to be digitized through the examination of official (2016) and Google Maps Street View (2016) was used to determine
trail maps or through a field investigation. Considering the latter the boundary points along each PSP that intersected with a road,
method when official trail maps were unavailable (Loyalsock designating the capability of access through entrance of a parking
Riverfront Trail and Muncy Heritage Nature Trail), Fitbit (2017) lot or other opening.
and GPS (global positioning system) smartphone technologies
were paired to map a digital representation of the walking path County Data
along the given trail. This data were then transferred into GIS line Because there was limited data available for individual household
format via digitization to complete trail line data inventory. data, county jurisdictions were contacted directly. Each county
The acquisition of park and pool data in the form of GIS digi- within the four-county region, as well as the adjacent counties that
tized polygon representation was also essential for the development fell within the project buffer boundaries (Centre, Columbia, Clinton,

© ASCE 04018030-5 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Dauphin, Juniata, Mifflin, Montour, Potter, Schuylkill, and Tioga), Trail Data
had either a GIS contact or department. The local metropolitan plan- To assess trail demand, both automatic and manual counting
ning organization (MPO), Susquehanna Economic Development methods were used. For the automatic counts, TRAFx (TRAFx
Association-Council of Governments (SEDA-COG), also assisted Research Ltd. 2016) infrared trail counters (passive) were installed
with gathering the household data for counties within its jurisdiction. every 8.05 km (5 mi) (if applicable) of the trail facility, reflective of
Once the data for each county were received, the following ap- the comfortable bikeshed distance. For six out of the seven trail
proach was taken in identifying residential homes for the study projects, only one counter was installed because the overall length
region. First, the address points layer was examined to see if it con- was less than 8.05 km (5 mi); but for one trail, two counters were
tained a structure type field. Address points that had any indication installed to more accurately count the trips over a distance further
that they were not residential, such as having a structure type code than the 8.05 km (5 mi) bike distance. The counters were installed
“BIZ” (business), were removed. Structures that fell into ambiguous on a mile marker or similar fixed object on point along the trail that
codes, such as “MIX” (mixed-use), or had no type information were was reflective of traffic flow. A minimum of 12 days of data were
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

kept. The address points were spatially joined to the land use or zon- collected at each location, then the data were downloaded from the
ing layer. Points that had land use or zoning descriptions indicating counter to the computer for analysis.
the point was not residential, such as “commercial,” were removed. For each counter that was installed in the field, a 2-h manual
Remaining points left with missing or ambiguous information were count was conducted to evaluate mode split, directional split, error
filtered using any additional information provided. Aerial imagery, adjustment, and other general characteristics about users (gender
Google Maps Street View (2016), and comment fields were used to and approximate age category). The manual counts were useful in
further remove nonresidential points. validating and determining the accuracy of the automatic counters.
Because the automatic equipment counts were based on a change in
Network Data infrared (temperature) as users walk by, when users walk in groups
A network data set containing the street network was necessary to or side-by-side, the accuracy can be affected. Therefore, manual
calculate the desired connectivity measures and create the project counts can be used to calculate an error adjustment factor by divid-
buffers needed for many of the metrics. The Network Analyst tools
ing the manual count total by the number recorded by the counter
in ArcGIS (ESRI 2016d) requires a street network data set. The
during that time (Ryus et al. 2014). Then, the error adjustment fac-
most current and accessible network available was a 2013 street
tor is applied to the automatic count data for a more accurate result.
network data set (ESRI 2013). The ArcGIS Online network data
In addition, because the counters do not distinguish between modes
set, which is updated regularly, was accessed in ArcGIS Desktop
or determine directional split, the additional information can be
using the Ready-To-Use-Service option. With this feature, the Gen-
useful in finding trends beyond just the number of users.
erate Service Areas tool was used to create the required 8.05 km
Another adjustment was used for linear trails because it was ex-
(5 mi) and 0.81 km (0.5 mi) network buffers around each project.
pected that users would be passing the counters twice (down and
This tool creates the network buffer around a point; thus, the access
back); therefore, the results were divided by two, also referred to as
points feature was initially used for the buffers. Later, the buffers
a two-way adjustment factor (TRAFx Research Ltd. 2016). For the
were dissolved into one buffer per pilot study project. Fig. 1 dis-
one linear trail with two counters installed, a “divide by 2.4” factor
plays the four-county region and the associated 77 PSP buffers. The
was used, based on the results of a previous study, which found that
8.05 km (5 mi) buffers were created using the default travel mode,
users passed the 1.0 and 8.5 mi counters an average of 2.4 times
which corresponds to driving; however, the half mile buffers were
during one trip (Oswald Beiler et al. 2012). However, for looped
created by setting custom walking attributes.
trails, this factor was not applied because it was assumed that trail
Miscellaneous Data users would typically complete the loop and exit the facility, pass-
To compute other metrics, additional data were needed beyond ing the counter once.
the project and county data. Additional recreation features such as Expansion factors were also applied to estimate the number of
fitness and recreational centers, boat access points, and preserved trips per year for each facility. For each facility, the average daily
areas were necessary in order to evaluate the metrics of non-PSP total count in the month of July was used to estimate the number of
density and the physical activity facility diversity index. Many of trips for the month of July. The National Bicycle and Pedestrian
these features were available on the online platform ESRI Business Documentation Project (NBPDP 2009) provides extrapolation fac-
Analyst by searching the physical activity establishments North tors to estimate the annual total trips from monthly data. The long
American Industry Classification System (NAICS) code (ESRI winter–short summer factor was used to expand the July totals,
2016b). In addition to Business Analyst, other recreation features which make up 13% of the annual totals, to the number of trips per
and physical activity establishments were downloaded from open year (NBPDP 2009). After the automatic data were downloaded,
source data repositories such as PASDA (2004, 2013, 2015). In the TRAFx DataNet software was used for the analysis process in
addition to recreation features, a population US Census (2010) data order to calculate trips per year (TRAFx Research Ltd. 2016) as
set at the MCD level was downloaded in order to compute metrics well as other results such as average hourly, weekly, and monthly
such as population density. counts. Trips per year became the input for the trail demand metric,
and the manual count results were used to determine the trail mode
split metric.
Field Data Collection (Step 5)
In order to evaluate the impact of recreational project development Pool Demand
on the community, exploring the demand of the facilities is valu- As a component of project usage analysis, current demand for the
able. In this study, both trail and pool facilities were investigated PSP pools within the pilot study was carried out through direct in-
because data could be acquired using manual and field data collec- quiry to appropriate pool representatives. Through online directory
tion equipment (trails) as well as contacting facility owners (pools). searches on borough or township websites and via direct calls to
Future work includes developing a process for gathering park borough offices, a representative was identified for each pool.
demand data given the challenge of multiple access points as well The representative was contacted with the request for a reported
as the multitude of park projects with varying amenities and features. average monthly attendance for the respective pool. The majority

© ASCE 04018030-6 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Five Mile Geodetic Buffer
Five Mile Network Buffer
DCNR Park
Park Access Point
Sample Household
Five Mile Road Network Route
Street Network
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Study area and 5-mi network project buffers. (GIS source data from PA DCNR 2014, 2015; ESRI 2013, 2016d; PASDA 2013, 2014a, b; RTI
International 2015; Centre County GIS 2015, 2016; Columbia County GIS 2015; Clinton County GIS 2006, 2015; Dauphin County GIS 2016;
Lycoming County GIS 2015; Juniata County GIS 2015; Mifflin County GIS 2010; Montour County GIS 2009; Northumberland County GIS
2015; Potter County GIS 2016; Schuylkill County GIS 2016; Snyder County GIS 2015; Tioga County GIS 2016; Union County GIS 2015.)

of representatives were willing to follow up with the solicited data For all metrics except those under the project demand category,
and were able to provide a form of monthly attendance. Assuming GIS was used in the analysis process. Table 3 indicates the specific
that the pool season spans from Memorial Day to Labor Day, all process and equations used for the metrics analyzed in GIS. Metrics
attendances were standardized in the units of people/season. This such as household density (shown at the MCD level in Fig. 2) and
value of attendees per season became the input for the pool demand PSP density involved creating a spatial join between the MCDs and
metric. buffers and the layer of interest, households, and PSP, respectively.
Similarly, the physical activity facility diversity index (shown at the
project buffer level in Fig. 3) and non-PSP density metrics involved
Metric Application (Step 6) joining the eight categories of physical activity establishments to
Once all relevant data were collected, the metrics were applied to the MCDs and buffers in order to sum the number of facilities fall-
the four-county pilot study region. The metrics in the categories of ing into the specific area. The connectivity measures involved tak-
density, connectivity, and physical activity facilities were applied ing various ratios of links, intersections, and dangle nodes. Links
and calculated at both the MCD level (123 MCDs) and the project were made by “unsplitting” the road network, intersections created
buffer level (77 project buffers). The metrics in the categories of by intersecting the road network with itself, and dangle nodes using
project demand and project magnitude were applied directly to the the “feature vertices to point” tool. Once these three features had
project and facility type. been extracted from the road network, measures like link-node ratio

© ASCE 04018030-7 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Table 3. GIS methodology for metric applications
Category Measure Calculation GIS method
Density Walkable household # of households per ½ mi Selection by location
density network distance buffer
Household density # of households per area Selection by location
Population density # of people per area Selection by location
Connectivity Roadway network Average network distance Network analyst: new OD cost matrix
distance from each household to (destinations = PSPs, origins = households)
nearest PSP access point
Geodetic distance to Euclidian distance from “Near” tool
project household to nearest PSP
access point
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Route directness Network distance/geodetic Field calculator


distance
Link-node ratio # Links (streets) per area/# Field calculator
nodes per area
Average block length Sum of link length per unit Spatial join (sum total length of streets in area) → spatial join
of area/# of nodes per unit (sum total nodes in area) → field calculator (divide)
of area
Intersection density # real nodes/area Spatial join (sum nodes), field calculator
Connected # real nodes/# total nodes Real nodes: (1) intersect (analysis): input features = road network:
node-ratio (CNR) (real + dangle) intersect with itself → (2) delete identical (data management):
removes superfluous instances
Dangle nodes: feature vertices to point type → dangle
Roadway density Total street length per unit Field calculator: Sum(Street_Length)/MCD_Area
of area/area
Physical activity facility Physical activity facility # of physical activity facility Spatial join (sum each of the eight individual physical activity
development diversity index types present in area/8 facility types).
Non-PSP density # Non-PSP physical activity Spatial join (sum each of the eight individual physical activity
facilities/area facility types). Sum all categories in excel and divide by area
PSP density # PSP/area spatial join (sum PSP), field calculator
Project magnitude Trail length Total length of trail Field calculator: Shape_Length
Pool facility area Total area of pool Field calculator: Shape_Area
Park facility area Total area of park Field calculator: Shape_Area
Note: Project demand metrics did not require GIS.

could be computed. Fig. 4 displays the link-node ratio for two dif- buffer than the 8.05 km (5 mi) buffer, suggesting that households
ferent MCDs, one with a low ratio and one with a high ratio, as well are generally clustered around the PSP. For some PSPs, the project
as showing the links, intersections, and dangle nodes. buffer-level and MCD-level application of the physical activity di-
Roadway network distance and geodetic distance to the project versity index both resulted in a value of 1.00, which means that
were the most complex GIS metrics. Both involved using Model- there is at least one establishment for all eight categories of physical
Builder to automate the complex process. Roadway network distance activity facility present at both levels. Surprisingly, some MCDs
required the use of Network Analyst in order to calculate the network have no physical activity facility, whereas within a project buffer,
distance from each household to the PSP access point. Fig. 5 shows the minimum is two types of facilities. Finally, it is interesting to
the difference between the roadway network distance and geodetic note the wide range of trips per year for the trail demand metric,
distance for a sample of households to a PSP park. from 880 to 95,000, as well as the wide range for pool demand,
from 7,300 to 27,000, reflective of the diversity of project scale
and usage.
Results In order to further analyze the results, statistical analysis was
For each metric, the GIS or field data collection application was performed using correlations between metric results. The metric
completed. The results are summarized in Table 4, showing the aver- correlations as well as the findings, specifically when comparing
age value as well as the range of values calculated for each metric. the project buffer-level and MCD-level application, are discussed
An interesting finding was that the project buffers, in general, have in depth.
greater area than the MCDs. Therefore, for measures such as road-
way network distance and geodetic distance, it makes sense that the Metric Correlation
average distance from household to project is greater for the project
buffer application than the MCD application. Although the average In order to evaluate the relationship between the transportation and
values for the roadway network distance and geodetic distance are land use metrics, a correlation table was created comparing each
lower for the MCD application, the route directness for the project metric to each other (except for project magnitude metrics, which
application is lower. This implies that although households may, on were only compared to project demand metrics due to correlation
average, be further from the PSP in the project buffers, the routes are relevance). By comparing each metric to each other, there is the
overall more direct. ability to determine correlations between project characteristics
Another interesting finding is that the average walkable house- and network accessibility. The Pearson product-moment correla-
hold density is larger than the average household density. The tion coefficient (also referred to as Pearson’s r) calculation was
household density, in general, is greater in the 0.81 km (0.5 mi) used to determine the statistical measurement of the correlation

© ASCE 04018030-8 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Household density application at the MCD level. (GIS source data from PASDA 2014a, b; US Census 2010; Centre County GIS 2015, 2016;
Columbia County GIS 2015; Clinton County GIS 2006, 2015; Dauphin County GIS 2016; Lycoming County GIS 2015; Juniata County GIS 2015;
Mifflin County GIS 2010; Montour County GIS 2009; Northumberland County GIS 2015; Potter County GIS 2016; Schuylkill County GIS 2016;
Snyder County GIS 2015; Tioga County GIS 2016; Union County GIS 2015.)

(linear association) between the two sets of metric values. The cor- the fact that the population surrounding a trail facility, based on
relation between the two sets of metric values was completed for the 8.05 km (5 mi) network buffer, has a direct impact on the
both project buffers and MCD jurisdictions. actual usage the trail receives. In contrast, Walkable Household
Table 5 displays the results of the metric correlation analysis. Density is not correlated with trail demand, suggesting trail
Within each cell, there are two correlations with the upper value usage is supported by a larger region (beyond the half mile
reflective of the project (project magnitude and project demand walking distance to the trail) and that trail users are willing to
metrics) or project buffer-level application and the lower value re- travel further to access a trail in comparison to other community
flective of the MCD-level application. Also, the first metric, area, facilities, such as a pool.
was added to the statistical analysis in order to explore relationships • Pool demand versus walkable household density—Even though
between project area, MCD area, and the transportation and land the population density within 8.05 km (5 mi) around a pool is
use metrics. Very strong positive correlations are indicated with a not an indicator for pool demand, the household density within
bold font (values from −1.0 to −0.8 and 0.8 to 1.0). Some specific 0.81 km (0.5 mi) walkable distance of the pool does strongly
highlights of the metric correlations are described subsequently: correlate to pool demand, suggesting that walkability for pools
• Population density versus trail demand—A very strong cor- is an indicator of pool usage.
relation between these two variables indicates that population • Intersection density versus population and household density—
density is a solid predictor for trail demand, and also reflects A strong positive correlation between intersection density and

© ASCE 04018030-9 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Physical activity facility diversity index applied at the project buffer level. Scale 0–1 based on presence of physical activity diversity types.
Project buffers are based on a 5-mi network buffer around the project. (GIS source data from PA DCNR 2014, 2015; ESRI 2013, 2016a; PASDA
2004, 2013, 2014a, b, 2015; Centre County GIS 2015, 2016; Columbia County GIS 2015; Clinton County GIS 2006, 2015; Dauphin County GIS
2016; Lycoming County GIS 2015; Juniata County GIS 2015; Mifflin County GIS 2010; Montour County GIS 2009; Northumberland County GIS
2015; Potter County GIS 2016; Schuylkill County GIS 2016; Snyder County GIS 2015; Tioga County GIS 2016; Union County GIS 2015.)

population and household density metrics validates the fact that metrics. Out of the three project types, pools are typically closer
areas with more intersections tend to be more compact in density in terms of both roadway network and geodetic distance.
and development. Fig. 6 displays the relationship of intersection Similar to Figs. 6 and 7, additional analysis of project type
density versus household density based on project type (park, (park, pool, and trail) can be explored for all metric correlations
pool, and trail). In this particular relationship, there is a strong as well as in relation to similar project characteristics (such as area,
positive correlation for all three project types, as represented by demand, etc.).
the fairly linear upward trend line. This relationship relates to pre- In terms of statistical significance, p-values were calculated
vious studies that confirm that these measures are interconnected for each correlation (both MCD and project level). Correlations
(Ewing and Cervero 2010). Ewing and Cervero (2010) further that were found to be statistically significant (less than 0.05)
suggested that intersection density is the single most important are indicated by an asterisk in Table 5. For the MCD level, a ma-
measure of walkability. jority of the correlations were statistically significant, excluding
• Roadway network distance versus geodetic distance—A strong those involving route directness and physical activity diversity
positive correlation between these two metrics confirms that index. For the project level, there were more correlations that
the roadway network reflects a similar “shortest pathway” ap- were not significant, suggesting that, in general, the larger scale
proach and avoids indirect routes within the project buffer. Fig. 7 at the MCD level provides more statistically significant results
displays the strong statistical correlation between these two between the transportation and land use metrics. However, it is

© ASCE 04018030-10 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Real Nodes Dangle Nodes Road Network Municipal Boundary
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

0 0.5 1 2
Miles

Fig. 4. Example of link-node ratio metric application. (GIS source data from ESRI 2013, 2016d; Centre County GIS 2015, 2016; Columbia County
GIS 2015; Clinton County GIS 2006, 2015; Dauphin County GIS 2016; Lycoming County GIS 2015; Juniata County GIS 2015; Mifflin County GIS
2010; Montour County GIS 2009; Northumberland County GIS 2015; Potter County GIS 2016; Schuylkill County GIS 2016; Snyder County GIS
2015; Tioga County GIS 2016; Union County GIS 2015.)

interesting to note that for the physical diversity index metric, a project buffers (low positive correlation of 0.282). Therefore,
majority of the correlations were less than 0.05, suggesting that non-PSPs are more likely to be observed in the same areas of
the smaller-scale project buffer level produced more significant PSPs with MCDs, but not as frequently in project buffers. This
results. For the demand metrics, pool demand had more statisti- relationship reflects the fact that geographic jurisdictions are
cally significant correlations than trail demand; however, future more officially organized through the jurisdiction of MCDs,
work includes gathering more data for trails and pools in order which have governmental organizations regulating the presence
to further investigate the relationships. of recreational facilities (similarities in funding efforts, organi-
zational commitment to recreational facilities, employee exper-
tise, etc.), whereas project buffers, without a similar oversight in
Project and MCD Comparison
planning or funding, contain a more random presence of both
In addition to comparing the metrics, exploring the correlations with PSPs and non-PSPs.
respect to the differences in the project buffer level versus the MCD • Link-node ratio versus roadway network distance and geodetic
level analysis is valuable. The following is a list of selected corre- distance—A strong negative correlation at the MCD level sug-
lations that varied between project and MCD, with a corresponding gests that people who live close to a project are more likely to
description suggesting the reasoning behind the difference. live in a connected street network. However, for the project buf-
• PSP density versus non-PSP density—A strong correlation be- fer level, there was no correlation, which is most likely a result
tween PSP density and non-PSP density for MCDs suggests of the project buffers being predefined based on a 8.05 km
that PSPs are located in MCDs that already have high densities (5 mi) network buffer from the project. Therefore, households
of other recreational facilities. However, this is not the case for included in the project analysis already reflect the nature of the

© ASCE 04018030-11 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Example of route directness metric application at the project buffer level. Project buffer is based on a 5-mi network buffer around the project. (GIS
source data from ESRI 2013, 2016d; Centre County GIS 2015, 2016; Columbia County GIS 2015; Clinton County GIS 2006, 2015; Dauphin County
GIS 2016; Lycoming County GIS 2015; Juniata County GIS 2015; Mifflin County GIS 2010; Montour County GIS 2009; Northumberland County GIS
2015; Potter County GIS 2016; Schuylkill County GIS 2016; Snyder County GIS 2015; Tioga County GIS 2016; Union County GIS 2015.)

street network, so variations between low and high connectivity surrounding these projects are more connected via dense road
are not as recognized within project buffers. infrastructure, producing a smaller average distance from house-
• MCD area versus roadway network and geodetic distance—A hold to project.
strong positive correlation between these variables reflects
the connection between jurisdictional size and traveling connec-
tivity. MCDs that have small jurisdictional area, some less than Connection to Public Health
1.61 km (1 sq mi), are are well connected and thus have lower
distance averages, whereas large MCDs tend to have more The focus of this pilot study research is evaluating transportation
sprawling populations and less road connectivity, leading to and recreational project development using metrics; however,
higher distance averages. In contrast, when analyzing the project understanding how health agencies and organizations can use
area, there is a weak negative correlation with the two distance the information in order to determine patterns can be explored
measures, which is again reflective of the fact that the project in preparation for connecting the process to public health measures.
buffer was created based on the network (street) distance of The Geisinger Health System region was used as part of the pilot
8.05 km (5 mi) around the project. The average distance to study because the four-county region falls within the Geisinger
the project will be higher for projects with smaller buffers be- service area. To expand the pilot study, future work can include
cause the road network surrounding this project is less direct. expanding the study to the entire 31-county service region using
However, for a project buffer with a larger area, the households the Geisinger electronic health record (EHR) data. The process

© ASCE 04018030-12 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Table 4. Summary of metric results


Metric Project buffers application (n ¼ 77) average (range) MCD application (n ¼ 123) average (range)

© ASCE
Area 68.25 (4.90–260.81) sq km 26.35 (1.89–100.70) sq mi 49.88 (0.26–206.6) sq km 19.26 (0.10–79.77) sq mi
Walkable household density 462.5 (0–1,793.98) no. per sq km 1,197.92 (0–4,646.40) no. per sq mi —
Household density 109.80 (1.62–280.16) no. per sq km 284.39 (4.20–725.61) no. per sq mi 128.25 (0.03–1,979.3) no. per sq km 332.18 (0.08–5,126.41) no. per sq mi
Population density 5,497.27 (1.05–171,338.78) no. per sq km 14,237.93 (2.73–44,367.44) no. per sq mi 286.54 (0.51–3,434.65) no. per sq km 742.15 (1.31–8,895.75) no. per sq mi
Roadway network distance 4.38 (2.33–6.81) km 2.72 (1.45–4.23) mi 2.45 (0.34–9.26) km 1.52 (0.21–5.74) mia
Geodetic distance 3.45 (1.67–5.06) km 2.14 (1.04–3.14) mi 1.77 (0.26–6.13) km 1.10 (0.16–3.81) mia
Route directness 1.27 (1.14–1.46) 1.38 (1.06–1.63)
Link-node ratio 1.49 (1.31–1.65) 1.42 (1.11–1.91)
Average block length 0.43 (0.24–1.96) km 0.27 (0.15–1.22) mi 0.61 (0.13–1.51) km 0.38 (0.08–0.94) mi
Intersection density 14.99 (2.66–36.63) no. per sq km 38.83 (6.88–94.88) no. per sq mi 16.32 (0.46–167.70) no. per sq km 42.28 (1.20–434.34) no. per sq mi
Connected node ratio 0.82 (0.69–0.89) 0.75 (0.53–0.95)
Roadway density 6.03 (3.33–11.21) km per sq km 9.70 (5.35–18.03) mi per sq mi 4.98 (0.90-39.90) km per sq km 8.01 (1.44–64.19) mi per sq mi
Physical activity facility 0.77 (0.25–1.00) 0.40 (0–1.00)
diversity indexb
Non-PSP density 0.67 (0.08–6.73) no. per sq km 1.73 (0.22–17.42) no. per sq mi 0.67 (0–5.49) no. per sq km 1.74 (0–14.22) no. per sq mi
PSP density 0.07 (0–0.21) no. per sq km 0.17 (0–0.54) no. per sq mi 0.135 (0–1.99) no. per sq km 0.35 (0–5.15) no. per sq mi
Trail demand 30,000 (880–95,000) no. trips per year —
Pool demand 13,000 (7,300–27,000) no. attendees per season —
Trail length 4.44 (0.58–15.39) km 2.76 (0.36–9.56) mi —
Pool facility area 10,683.71 (6,191.7–16,106.50) sq m 2.64 (1.53–3.98) ac. —
Park facility area 80,896.73 (849.84– 1,701,987.91) sq m 19.99 (0.21–420.57) ac. —
a
Only including MCDs with PSP (n ¼ 55).
b
Scale: (0 to 1) based on presence of eight physical activity facility types.

04018030-13

J. Urban Plann. Dev., 2018, 144(4): 04018030


J. Urban Plann. Dev.
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Table 5. Metric correlation analysis for project/project buffer and MCD level application
Physical

© ASCE
Walkable Roadway Average Connected activity Pool
household Household Population network Geodetic Route Link-node block Intersection node Roadway diversity Non-PSP PSP Trail Pool Trail facility
Metrics Area density density density distance distance directness ratio length density ratio density index density density demand demand length area
Areaa 1.00 0.058 −0.001 0.376b −0.252b −0.173 −0.237b −0.152 −0.371b 0.016 0.017 −0.322b 0.377b −0.209 0.053 0.482 −0.702 — —
1.00 — −0.409b −0.444b 0.868b 0.874b 0.269b 0.657b 0.610b −0.479b −0.628b −0.524b 0.162 −0.466b −0.409b — — — —
Walkable household — 1.00 0.522b 0.342b −0.595 −0.520b −0.285b 0.491b −0.457b 0.489b 0.490b 0.253 0.157 −0.058 0.260b 0.063 0.866b — —
density 1.00 — — — — — — — — — — — — — — — — —
Household density — 1.00 0.835b −0.119 −0.149 0.069 0.767b −0.625b 0.991b 0.790b 0.767 0.462b 0.273b 0.853b 0.533 0.715b — —
1.00 0.969b −0.469b −0.486b −0.184 0.647b −0.501b 0.937b 0.577b 0.818b 0.039 0.630b 0.637b — — — —
Population density — — — 1.00 −0.154b −0.170 0.020 0.559b −0.576b 0.859b 0.669b 0.603 0.632b 0.288b 0.816b 0.917b −0.006 — —
1.00 −0.528b −0.548b −0.181 0.675b −0.547b 0.950b 0.610b 0.835b 0.112 0.736b 0.664b — — — —
Roadway network — — — — 1.00 0.936b 0.327b 0.023 0.364b −0.120 −0.084b 0.089 −0.283b −0.001 −0.005 −0.385 −0.769 — —
distance 1.00 0.990b 0.412b −0.773b 0.799b −0.581b −0.779b −0.671b 0.092 −0.584b −0.571b — — — —
Geodetic distance — — — — — 1.00 −0.020 −0.041 0.298b −0.164 −0.128 −0.012 −0.286b −0.063 −0.013 −0.431 −0.760b — —
1.00 0.307b −0.803b 0.791b −0.603b −0.806b −0.699b 0.110 −0.612b −0.598b — — — —
Route directness — — — — — — 1.00 0.175 0.213 0.105 0.116 0.271b −0.043 0.163 0.022 −0.120 −0.200 — —
1.00 −0.224 0.293b −0.204 −0.270b −0.204 0.100 −0.080 −0.175 — — — —
Link-node ratio — — — — — — — 1.00 −0.289b 0.753b 0.935b 0.647 0.232b 0.058 0.483b −0.136 0.748b — —
1.00 −0.493b 0.733b 0.924b 0.789b −0.104 0.644b 0.628b — — — —
Average block — — — — — — — — 1.00 −0.620b −0.359b −0.167 −0.532b 0.029 — −0.438 −0.508 — —
length 1.00 −0.582b −0.531b −0.581b −0.249b −0.563b 0.551b — — — —
−0.475b
Intersection density — — — — — — — — — 1.00 0.785b 0.792 0.508b 0.305b 0.857b 0.498 0.641 — —
1.00 0.650b 0.924b 0.068 0.759b 0.734b — — — —
Connected node — — — — — — — — — — 1.00 0.577 0.360b −0.015 0.554b 0.037 0.720b — —

04018030-14
ratio 1.00 0.678b −0.039 0.575b 0.537b — — — —
Roadway density — — — — — — — — — — — 1.00 0.237b 0.544b 0.631b 0.298 0.656 — —
1.00 −0.034 0.686b 0.668b — — — —
Physical activity — — — — — — — — — — — — 1.00 0.153 0.537b 0.515 −0.398 — —
diversity index 1.00 0.200b 0.012 — — — —
Non-PSP density 1.00

J. Urban Plann. Dev., 2018, 144(4): 04018030


— — — — — — — — — — — — — 0.282b 0.334 −0.088 — —
1.00 0.836b — — — —
PSP density — — — — — — — — — — — — — — 1.00 0.602 0.388 — —
1.00 — — — —
Trail demand — — — — — — — — — — — — — — — 1.00 — 0.690b —
1.00 — — —
Pool demand — — — — — — — — — — — — — — — — 1.00 — −0.281
1.00 — —
Trail length — — — — — — — — — — — — — — — — — 1.00 —
1.00 —
Pool facility area — — — — — — — — — — — — — — — — — 1.00
1.00
Note: Within each cell, the upper row is either the project or the project buffer level results, whereas the lower row is the MCD level results; bold font indicates very strong correlation (−1.0 to −0.8 or 0.8 to 1.0);
park facility area metric was not included in the table because park demand data were not part of the pilot study (no correlations).
a
Area metric was added to analyze project buffer area and MCD area in relation to the transportation and land use metrics.
b
p-value is less than 0.05, which indicates it is statistically significant (if there is no superscript, then p-value is more than 0.05 and is not statistically significant).

J. Urban Plann. Dev.


versus an area without and compare rates of cardiovascular
diseases.
The steps completed so far serve as the foundation for estab-
lishing the connection with health agencies and quantifying land
use development. Due to the scope of this pilot study, as well as
limitations associated with accessing Geisinger’s EHR data, com-
prehensive connections are identified as future work; however, the
following section discusses a preliminary connection to public
health using publically accessible data. This preliminary connec-
tion at the aggregate level should be used in terms of the example
process that can be applied once the Geisinger EHR data are ac-
quired (or any other direct health care system dataset).
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Preliminary Connection
A preliminary connection was made between the transportation and
Fig. 6. Intersection density versus household density by project type.
land use metrics and an open-source 2015 neighborhood map of US
obesity rates (RTI International 2015). Given the year of release, all
DCNR projects would have been completed before the data were
collected for the open-source maps. Obesity rates were provided at
the neighborhood level and were first aggregated to the MCD level
in the form of an average rate (Fig. 8). Similarly, the obesity rates
were also aggregated to the project buffer level.
Performing correlational analysis at both the project buffer and
MCD levels returned two interesting relationships. Observing a
−0.61 correlation between non-PSP density and average obesity
rate metrics at the project buffer level provides insight into a po-
tential connection between the presence of recreational facilities
and average obesity rates in the surrounding service area. The
p-value for this correlation was less than 0.05, indicating that it
is statistically significant. Another potential relationship observed
was that of a −0.39 correlation between trail demand and average
obesity rates at the project buffer level, perhaps reflecting the return
of trail infrastructure on increasing public health. These two cor-
Fig. 7. Roadway network distance versus geodetic distance. relations were both at the project buffer level, whereas the corre-
lations between MCD transportation metrics and average obesity
rates were weak. This may highlight the importance of installing
trail infrastructure in areas where public health is at a low because
for connecting the pilot study process to public health metrics using the findings suggest that trails most directly impact the public
EHRs as well as more accessible publically available data are dis- health of the populations immediately surrounding them, within
cussed below. a 8.05 km (5 mi) network buffer.
Further analysis using the p-value significance test shows that at
the MCD and project buffer levels, metric correlations such as aver-
Process age block length, connected node ratio, and link node ratio are stat-
Once the transportation and land use metrics have been applied, istically significant when compared to average obesity rates. At the
there is the opportunity for connecting with a public health agency. project level, the non-PSP density metric is also statistically signifi-
For the pilot study, the Geisinger Health System is the established cant, which suggests that when looking at the broader number
primary health service for the four-county region. Using the clinic’s of physical activity facilities at the smaller project level, there is
electronic health record data (EHR) (individual patient health re- a significance in the relationship to average obesity. Again, the
cords), public health metrics can be aggregated spatially. Metrics exploration of confounding variables is recommended when com-
such as body mass index (BMI), incidences of cardiovascular dis- pleting a full investigation of correlations between health and land
eases, and childhood obesity rates can be gathered for all patients use data (Jarrett et al. 2010).
within the study area within the EHR data. Any confounding var-
iables should be identified up front and minimized as completed in
Benefits
past studies (Jarrett et al. 2010). Then, spatial as well as temporal
comparisons can be made, as listed subsequently: The process developed is one that can be conducted by transpor-
• Temporal—Compare patient health before and after project tation planning agencies in connection with public health organi-
development. For example, identifying a high-demand trail fa- zations in order to compare development factors such as density,
cility and exploring average BMI pre-trail development versus connectivity, physical activity facilities, project demand, and
average BMI post-trail development. project magnitude with health related measures. This increased col-
• Spatial—Compare patient health for those living near project laboration can allow for a robust understanding of how a local com-
development or other physical activity facilities, as well as those munity is developed and how land use decisions can influence
within a connected network. For example, determine an area health and wellness. For example, the results of this study show
with a high density and diversity of physical activity facilities that walkability to a pool facility has a high correlation to demand.

© ASCE 04018030-15 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Percent Obese Population
Obesity Rate (Frequency)
27.68 (1)
27.69 - 37.56 (16)
37.57 - 38.42 (18)
38.43 - 39.00 (20)
39.01 - 40.05 (15)
40.06 - 41.18 (5)
41.19 - 48.05 (2)
Area with data avaliable
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Obesity rates at the project buffer level. Project buffers are based on a 5-mi network buffer around the project. (GIS source data from PA DCNR,
2014, 2015; ESRI 2013, 2016d; PASDA, 2013, 2014a, b; RTI International 2015; Centre County GIS 2015, 2016; Columbia County GIS 2015; Clinton
County GIS 2006, 2015; Dauphin County GIS 2016; Lycoming County GIS 2015; Juniata County GIS 2015; Mifflin County GIS 2010; Montour
County GIS 2009; Northumberland County GIS 2015; Potter County GIS 2016; Schuylkill County GIS 2016; Snyder County GIS 2015; Tioga County
GIS 2016; Union County GIS 2015.)

A public health agency could then further explore the rate of child- research include specific limitations throughout the study as well
hood obesity in areas with and without pools. Also, a local planning as recommendations on future work.
agency can identify priority locations for pool facilities and other
recreational projects. Understanding the relationship between the Limitations
built environment and public health and wellness can better inform
future urban planning decision-making processes. The following list includes highlights of the limitations and as-
sumptions made, many of which have already been described
previously.
Discussion 1. Project data were based on the PA DCNR grant database and
GIS shapefile; therefore, any missing information (or projects
The methodology used and results obtained in the pilot study outside of the study timeframe or scope) was not included in
application are valuable in providing a global process for explor- the study [see “Project Inventory (Step 2)”]
ing transportation and recreational projects. Transportation and 2. Trail demand was based on a minimum of 2 weeks of automatic
land use agencies across the county can use this pilot study as an counts and one 2-h manual count for adjustments. Therefore,
example of how to quantify development patterns across multiple yearly trip rates are based on seasonal adjustment factors for
jurisdictions and allow for the opportunity to connect with health long winter-short summer geographic regions. Also, linear trails
care organizations in their area. The following reflections on the were analyzed using a two-way adjustment factor of dividing by

© ASCE 04018030-16 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


2, or 2.4 for the trail longer than 8.05 km (5 mi), to account for • Explore similar characteristics of projects (including demand
users passing the counters multiple times (converts counter hits and scale) across statistically significant correlations to identify
to trips). any additional trends in project impacts.
3. Project network buffers were based on a 8.05 km (5 mi) network
distance from the project access point(s), and, in order to avoid
the modified areal unit problem (analyzing spatial data based on Conclusion
predetermined boundaries such as county boundaries), adjacent
counties outside of the four-county region were included for a Bridging the gap between development decisions and the impact
more comprehensive analysis of the project buffers. on human health is valuable because efforts to promote sustainable
4. Pool demand data were based on the information provided by community development continue. Transportation and recreational
the pool owners or operators (some facilities provided raw data, projects such as parks, pools, and trails provide the opportunity for
whereas others provided a general value for average attendees physical activity. Understanding patterns in the development of
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

per day). Future work includes reassessing the pool demand by these facilities, including measures such as demand, density, and
having all pool owners or operators provide raw data for an en- connectivity, can aid in future planning and funding.
tire season in order to determine the demand. This research included two primary goals: (1) develop a process
5. The following assumptions were made with regard to geo- for analyzing sustainable transportation and recreation projects us-
graphic information systems and the spatial analysis component ing project-specific and network-specific metrics; and (2) imple-
of this research: ment the process through a pilot study in central Pennsylvania.
a. The residential designation for households across the 14 The pilot study showed that trail demand was highly correlated to
counties was determined based on existing structure type population density within the project buffer, a 8.05 km (5 mi) net-
and land use codes as well as a visual assessment of Google work distance from the project, whereas pool demand was highly
Maps Street View (2016) and Google Earth (2016). Based on correlated to the walkable household distance of 0.81 km (0.5 mile)
the GIS files provided by the local sources, some of the at- network distance from the project. At the MCD level, MCDs with a
tribute fields were left blank or given ambiguous codes; thus, high density of PSPs also had a high density of non-PSPs, whereas
judgment was used when determining which buildings to in- MCDs with a high link-node ratio corresponded with those that
clude as households. had a lower average roadway network distance from project to
b. Access points for projects were determined based on visual household. A preliminary connection to public health metrics using
assessment whenever the road network data set connected publically accessible data (obesity rates) was used. Future work in-
with the facility boundaries. This assumption does not ac- cludes expanding the analysis to health measures both spatially
count for various other pedestrian paths or shortcuts that may (trends in patient condition in reference to household and project
exist. locations) and temporally (trends in patient condition over time be-
c. Both ArcGIS Online and ArcGIS Desktop street networks fore and after projects are developed) using individual health record
were used for different metric calculations due to the avail- data. Understanding this relationship between the built environ-
ability of GIS tools for metric applications. ment (specifically recreational facilities) and health can further sup-
port the goal of building sustainable and healthy communities.

Future Work
Because this is a pilot study focused on transportation and land use Acknowledgments
metrics, recommendations are provided for future work. Future work The authors would like to thank the Bucknell Geisinger Research
includes implementing the connection to public health measures with Initiative (BGRI) for supporting this research, as well as the follow-
the Geisinger electronic health record. This connection, described ing organizations for their efforts in support of the study: PA DCNR,
previously, allows for an understanding of the development associ- SEDA-COG GIS Department, Geisinger Health System, PA Bureau
ations on public health, specifically with regard to transportation and of Forestry, owners and employees of the trail and pool projects
recreational projects. The following future work is recommended included in the study, as well as the following county GIS depart-
based on the elements completed in the pilot study: ments: Centre, Columbia, Clinton, Dauphin, Lycoming, Juniata,
• Expand the study to the entire 31-county Geisinger Health Sys- Mifflin, Montour, Northumberland, Potter, Schuylkill, Snyder,
tem region as well as exploring applications in other geographical Tioga, and Union. With respect to all counties, these maps were
regions. created using county geographic information system digital data,
• Include travel time metrics in addition to the distance mea- but this is a secondary product and has not been verified and is
sures already explored in this study. This would be particularly not authorized by the individual counties themselves.
valuable when analyzing an urban area where distance and
time are not necessarily similar due to traffic volumes and
congestion. References
• Complete a demand analysis of the park projects using a com-
prehensive method for determining users entering and exiting Adams, E., F. Bull, and C. Foster. 2016. “Are perceptions of the environ-
at multiple access points. This would complete the data for the ment in the workplace ‘neighbourhood’ associated with commuter
project demand metrics. Also, a comparison of the amenities walking?” J. Transp. Health, 3 (4): 479–484. https://doi.org/10.1016/j
provided at the park facilities could be valuable because the .jth.2016.01.001.
Appleyard, B. 2012. “Sustainable and healthy travel choices and built envi-
park projects widely varied based on form and function (e.g., a
ronment.” Transp. Res. Rec. 2303 (1): 38–45. https://doi.org/10.3141
facility with tennis courts versus a nature park). /2303-05.
• Perform a comprehensive cluster analysis based on similar pro- Bakalar, N. 2016. “Walkable neighborhoods cut obesity and diabetes rates.”
ject types for all metrics to determine if there are any trends The New York Times. Accessed June 10, 2016. http://well.blogs.nytimes
when comparing park, pool, and trail projects in relation to the .com/2016/05/24/walkable-neighborhoods-cut-obesity-and-diabetes
five categories of metrics. -rates/?r=0.

© ASCE 04018030-17 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Braun, L., D. Rodriguez, Y. Song, K. Meyer, C. Lewis, J. Reis, and P. ESRI (Environmental Systems Research Institute). 2016b. “Travel modes.”
Gordon-Larson. 2016. “Changes in walking, body mass index and Accessed July 1, 2016. https://doc.arcgis.com/en/arcgisonline/reference
cardiometabolic risk factors following residential location: Longitudinal /travel-modes.htm.
results from the CARDIA Study.” J. Transp. Health 3 (4): 426–439. ESRI (Environmental Systems Research Institute). 2016c. “What is
https://doi.org/10.1016/j.jth.2016.08.006. ModelBuilder?” Accessed July 1, 2016. http://desktop.arcgis.com/en
Buehler, R., T. Götschi, and M. Winters. 2016. “Moving toward active trans- /arcmap/latest/analyze/modelbuilder/what-is-modelbuilder.htm.
portation: How policies can encourage walking and bicycling.” Active ESRI (Environmental Systems Research Institute). 2016d. “What is the
Living Research. Accessed June 10, 2016. http://activelivingresearch ArcGIS network analyst extension?” Accessed July 1, 2016. http://
.org/sites/default/files/ALR_Review_ActiveTransport_January2016.pdf. desktop.arcgis.com/en/arcmap/latest/extensions/network-analyst/what-is
Burbidge, S. K., and K. G. Goulias. 2009. “Evaluating the impact of neigh- -network-analyst-.htm.
borhood trail development on active travel behavior and overall physical Ewing, R., and R. Cervero. 2010. “Travel and the built environment.”
activity of suburban residents.” Transp. Res. Rec. 2135 (1): 78–86. J. Am. Plann. Assoc. 76 (3): 265–294. https://doi.org/10.1080
https://doi.org/10.3141/2135-10. /01944361003766766.
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

Cannon, C., S. Thomas, R. Treffers, M. Paschall, L. Heumann, G. Mann, D. Falconer, C. L., S. D. Leary, A. S. Page, and A. R. Cooper. 2015. “The
Dunkell, and S. Nauenberg. 2013. “Testing the results of municipal tracking of active travel and its relationship with body composition
mixed-use zoning ordinances.” Robert Wood Johnson Foundation. in UK adolescents.” J. Transp. Health 2 (4): 483–489. https://doi
Accessed October 17, 2017. https://www.rwjf.org/en/library/research .org/10.1016/j.jth.2015.09.005.
/2013/05/testing-the-results-of-municipal-mixed-use-zoning-ordinances FHWA (Federal Highway Administration). 2011. “Pedestrian and bi-
.html. cycle data collection.” US Dept. of Transportation, Federal Highway
CDC (Centers for Disease Control and Prevention). 1999. “Physical activ- Administration. Accessed July 5, 2016. https://www.fhwa.dot.gov
ity and health: A report from the surgeon general.” Accessed October /policyinformation/travel_monitoring/pubs/pedbikedata.cfm.
17, 2017. https://www.cdc.gov/nccdphp/sgr/index.htm. Fitbit. 2017. “Fitbit.” Accessed July 15, 2016. http://www.fitbit.com.
CDC (Centers for Disease Control and Prevention). 2010. “CDC recom- Forsyth, A., and J. M. Oakes. 2014. “Workplace neighborhoods, walking,
mendations for improving health through transportation policy.” physical activity, weight status, and perceived health.” Transp. Res. Rec.
Accessed June 14, 2016. https://www.cdc.gov/transportation/docs 2452 (1): 98–104. https://doi.org/10.3141/2452-12.
/final-cdc-transportation-recommendations-4-28-2010.pdf. FTA (Federal Transit Administration). 2011. “Final policy statement on the
CDC (Centers for Disease Control and Prevention). 2015. “Data trends eligibility of pedestrian and bicycle improvements under federal transit
and maps.” Accessed June 16, 2016. https://www.cdc.gov/obesity/data law.” Fed. Reg. 76 (161): 52046–52053.
/databases.html. Google Earth. 2016. “Central Pennsylvania.” Accessed June 15, 2016.
Center for Rural Pennsylvania. 2014. “Pennsylvania local parks and the https://www.google.com/earth/.
communities they serve.” Accessed June 16, 2016. http://www.paoutdoor Google Maps Street View. 2016. “Central Pennsylvania.” Accessed June 15,
recplan.com/cs/groups/public/documents/document/dcnr_20030877.pdf. 2016. https://www.google.com/maps/streetview/.
Centre County GIS. 2015. Buildings, inactive parcels, unofficial county Gori, S., M. Nigro, and M. Petrelli. 2014. “Walkability indicators for
zoning, GIS shapefile. Lewisburg, PA: SEDA-Council of Governments. pedestrian-friendly design.” Transp. Res. Rec. 2464 (1): 38–45. https://
Centre County GIS. 2016. Parcels, GIS shapefile. Lewisburg, PA: SEDA- doi.org/10.3141/2464-05.
Council of Governments. Harnik, P., and B. Welle. 2011. “From fitness zones to the medical mile:
Clinton County GIS. 2006. Zoning, GIS shapefile. Lewisburg, PA: SEDA- How urban park systems can best promote health and wellness.” The
Council of Governments. Trust for Public Land. Accessed June 15, 2016. http://cloud.tpl.org
Clinton County GIS. 2015. Buildings, parcels, structures, GIS shapefile. /pubs/ccpe-health-promoting-parks-rpt.pdf.
Lewisburg, PA: SEDA-Council of Governments. Heinen, E., J. Panter, A. Dalton, A. Jones, and D. Ogilvie. 2015. “Socio-
Cohen, D. A., D. Golinelli, S. Williamson, A. Sehgal, T. Marsh, and T. L. spatial patterning of the use of new transport infrastructure: Walking,
McKenzie. 2009. “Effects of park improvements on park use and physi- cycling and bus travel on the Cambridgeshire guided busway.”
cal activity: Policy and programming implication.” Am. J. Preventive J. Transp. Health 2 (2): 199–211. https://doi.org/10.1016/j.jth.2014
Med. 37 (6): 475–480. https://doi.org/10.1016/j.amepre.2009.07.017. .10.006.
Columbia County GIS. 2015. Parcels, structures, zoning, GIS shapefile. Iseki, H., and M. Tingstrom. 2013. “A new approach in the GIS bikeshed
Lewisburg, PA: SEDA-Council of Governments. analysis with consideration of topography, street connectivity, and
Dalton, A. M., A. P. Jones, J. Panter, and D. Ogilvie. 2014. “Are energy consumption.” In Proc., Transportation Research Board Annual
GIS-modelled routes a useful proxy for the actual routes followed Meeting 2013. Washington, DC: Transportation Research Board.
by commuters?” J. Transp. Health 2 (2): 219–229. https://doi.org/10 Jarrett, M., S. Gale, and C. Kontgis. 2010. “Spatial modeling in environ-
.1016/j.jth.2014.10.001. mental and public health research.” Int. J. Environ. Res. Public Health
Dannenberg, A. L., A. Ricklin, C. Ross, M. Schwartz, J. West, S. White, 7 (4): 1302–1329. https://doi.org/10.3390/ijerph7041302.
and M. Wier. 2014. “Use of health impact assessment for transportation Juniata County GIS. 2015. Parcels, structures, GIS shapefile. Lewisburg,
planning.” Transp. Res. Rec. 2452 (1): 71–80. https://doi.org/10.3141 PA: SEDA-Council of Governments.
/2452-09. Litman, T. 2003. “Economic value of walkability.” Transp. Res. Rec.
Dauphin County GIS. 2016. Address points, land use, GIS shapefile. 1828: 3–11. https://doi.org/10.3141/1828-01.
Harrisburg, PA: Dauphin County Information Technology Dept. Lycoming County GIS. 2015. Address points, parcels, GIS shapefile.
Dill, J. 2004. “Measuring network connectivity for bicycling and walking.” Williamsport, PA: Lycoming County Planning Dept.
Presented at the Active Living Research Annual Conference, Del Mar, Meloni, I., A. Portoghese, M. Bez, and E. Spissu. 2009. “Effects of physical
CA. Accessed May 17, 2016. http://reconnectingamerica.org/assets activity on propensity for sustainable trips.” Transp. Res. Rec. 2134 (1):
/Uploads/TRB2004-001550.pdf. 43–50. https://doi.org/10.3141/2134-06.
Ervin, D. 2016. “An introduction to the modifiable areal unit problem.” Mifflin County GIS. 2010. Buildings, land use, GIS shapefile. Lewistown,
Advanced Spatial Analysis. Accessed July 1, 2016. http://gispopsci PA: Mifflin County GIS Dept.
.org/maup/. Montour County GIS. 2009. Address points, land use, zoning, parcels, GIS
ESRI (Environmental Systems Research Institute). 2013. “Streets, ESRI shapefile. Mechanicsville, PA: Montour County Planning Commission.
data and maps.” Accessed July 1, 2016. http://www.esri.com/data NBPDP (National Bicycle and Pedestrian Documentation Project). 2009.
/data-maps. NBPD fact sheet and status report. Portland, OR: Alta Planning and
ESRI (Environmental Systems Research Institute). 2016a. “Business Design.
analyst.” Accessed July 1, 2016. http://www.esri.com/software/business Northumberland County GIS. 2015. Structures, parcels, GIS shapefile.
analyst. Sunbury, PA: Northumberland County Dept. of Information Services.

© ASCE 04018030-18 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030


Oswald Beiler, M., K. Burkhart, and M. Nicholson. 2015. “Evaluating Schlossberg, M., and C. Brehm. 2009. “Participatory geographic informa-
the impact of rail-trails: A methodology for assessing travel demand tion systems and active transportation.” Transp. Res. Rec. 2105 (1):
and economic impacts.” Int. J. Sustainable Transp. 9 (7): 509–519. 83–91. https://doi.org/10.3141/2105-11.
https://doi.org/10.1080/15568318.2013.825035. Schuylkill County GIS. 2016. Address points, parcels, GIS shapefile.
Oswald Beiler, M., T. Kinnaman, K. Burkhart, and M. Nicholson. 2012. Pottsville, PA: Schuylkill County Office of Geographical Information.
“Buffalo Valley rail trail 2012 user survey and economic impact assess- Schwartz, B. S., W. F. Stewart, S. Godby, J. Pollak, J. DeWalle, S. Larson,
ment.” Prepared for Union County, PA and BVRT Committee. Accessed D. G. Mercer, and T. A. Glass. 2011. “Body mass index and the built
June 16, 2016. http://bvrt.org/2012-BVRTUserSurveyEconomicImpact and social environments in children and adolescents using electronic
Analysis.pdf. health records.” Am. J. Preventive Med. 41 (4): e17–e28. https://doi
PA DCNR (Pennsylvania Department of Conservation and Natural Resour- .org/10.1016/j.amepre.2011.06.038.
ces). 2014. Local parks, GIS shapefile, DCNR open data. Harrisburg, Semler, C., A. Vest, K. Kingsley, S. Mah, W. Kittelson, C. Sundstrom,
PA: PA DCNR.
and K. Brookshire. 2016. “Guidebook for developing pedestrian and
PA DCNR (Pennsylvania Department of Conservation and Natural Resour-
Downloaded from ascelibrary.org by Universidad Militar Nueva Granada on 03/18/19. Copyright ASCE. For personal use only; all rights reserved.

bicycle performance measures.” US Dept. of Transportation, Federal


ces). 2015. Pennsylvania DCNR bureau of recreation and conservation
Highway Administration. Accessed June 15, 2016. https://www
projects funded through C2P2, GIS shapefile. Harrisburg, PA: PA
.fhwa.dot.gov/environment/bicycle_pedestrian/publications/performance
DCNR.
PA DCNR (Pennsylvania Department of Conservation and Natural Resour- _measures_guidebook/.
ces). 2016. “Grants.” Accessed June 15, 2016. https://www.grants.dcnr Silver, D., M. Giorgio, and T. Mijanovich. 2014. “Utilization patterns and
.state.pa.us/Dashboard/Grants. perceptions of playground users in New York City.” J. Community
PA DCNR (Pennsylvania Department of Conservation and Natural Resour- Health 39 (2): 363–371. https://doi.org/10.1007/s10900-013-9771-0.
ces). 2017. “Trail grants.” Accessed October 17, 2017. http://www.dcnr Snyder County GIS. 2015. Address points, land use, parcels, GIS shapefile.
.pa.gov/Communities/Grants/TrailGrants/Pages/default.aspx. Lewisburg, PA: Union County GIS Dept.
PASDA (Pennsylvania Spatial Data Access). 2004. “Protected lands Stewart, O., H. Carlos, C. Lee, E. Berke, P. Hurvitz, L. Li, A. Moudon,
inventory: Federal land, GIS shapefile, the conservation fund.” Ac- and M. Doescher. 2016. “Secondary GIS built environment data for
cessed June 15, 2016. http://www.pasda.psu.edu/uci/DataSummary health research: Guidance for data development.” J. Transp. Health
.aspx?dataset=884. 3 (4): 529–539. https://doi.org/10.1016/j.jth.2015.12.003.
PASDA (Pennsylvania Spatial Data Access). 2013. “Explore PA trails: Tioga County GIS. 2016. Parcels, structures, GIS shapefile. Wellsboro,
Trails (points), GIS Shapefile, PA DCNR.” Accessed June 17, 2016. PA: Tioga County GIS Dept.
http://www.pasda.psu.edu/uci/DataSummary.aspx?dataset=1551. TRAFx Research Ltd. 2016. “Trail infrared counters.” Accessed June 1,
PASDA (Pennsylvania Spatial Data Access). 2014a. “County boundaries, 2016. https://www.trafx.net/products.htm#IR-counter.
GIS shapefile.” Accessed June 15, 2016. http://www.pasda.psu.edu/uci Tresidder, M. 2005. “Using GIS to measure connectivity: An exploration
/DataSummary.aspx?dataset=24. of issues.” Portland State Univ., School of Urban Studies and Planning.
PASDA (Pennsylvania Spatial Data Access). 2014b. “Municipal bounda- Accessed June 1, 2016. http://www.web.pdx.edu/∼jdill/Tresidder_Using
ries, GIS shapefile.” Accessed June 15, 2016. http://www.pasda.psu.edu _GIS_to_Measure_Connectivity.pdf.
/uci/DataSummary.aspx?dataset=41. Union County GIS. 2015. Parcels, address points, GIS shapefile.
PASDA (Pennsylvania Spatial Data Access). 2015. “Access points (fishing Lewisburg, PA: Union County GIS Dept.
and boating), GIS shapefile.” Accessed June 15, 2016. http://www Urban Land Institute. 2015. “Building healthy places toolkit: Strategies for
.pasda.psu.edu/uci/DataSummary.aspx?dataset=984.
enhancing health in the built environment.” Accessed August 15, 2016.
Potter County GIS. 2016. Parcels, structures, GIS shapefile. Coudersport,
http://uli.org/wp-content/uploads/ULIDocuments/Building-Healthy-Places
PA: Potter County Dept. of GIS/Planning.
-Toolkit.pdf.
RTI International. 2015. “US obesity data, GIS shapefile, synthetic micro-
US Census. 2010. “TIGER/line with selected demographic and economic
data.” Accessed July 1, 2016. http://synthpopviewer.rti.org/obesity
/download.html. data.” Demographic Profile 1 County Subdivisions, GIS Shapefile,
Ryus, P., E. Ferguson, K. M. Laustsen, R. J. Schneider, F. R. Proulx, T. 2010 Census. Accessed June 14, 2016. https://www.census.gov/geo
Hull, and L. Miranda-Moreno. 2014. Guidebook on pedestrian and /maps-data/data/tiger-data.html.
bicycle volume data collection. NCHRP Rep. 797. Washington, DC: USDOT (US Department of Transportation). 2016. “Transportation and
Transportation Research Board. health indicators.” Accessed June 14, 2016. https://www.transportation
Samimi, A., and A. K. Mohammadian. 2010. “Health impacts of urban de- .gov/transportation-health-tool/indicators.
velopment and transportation systems.” J. Urban Plann. Dev. 136 (3): WHO. 2014. “Health and economic assessment tool for cycling and walk-
208–213. https://doi.org/10.1061/(ASCE)UP.1943-5444.0000020. ing.” Accessed June 16, 2016. http://www.heatwalkingcycling.org/.
Schauder, S., and M. Foley. 2015. “The relationship between active trans- Zhang, X., J. Lu, and J. B. Holt. 2011. “Modeling spatial accessibility to
portation and health.” J. Transp. Health 2 (3): 343–349. https://doi.org parks: A national study.” Int. J. Health Geogr. 10 (1): 31. https://doi.org
/10.1016/j.jth.2015.06.006. /10.1186/1476-072X-10-31.

© ASCE 04018030-19 J. Urban Plann. Dev.

J. Urban Plann. Dev., 2018, 144(4): 04018030

You might also like