The impact of small changes in thoroughfare connectivity on the potential for student walking
TThe impact of small changes in thoroughfare connectivity on the potential for student walking Jeremy D. Auerbach , Eugene C. Fitzhugh , and Ellen Zavisca Department of Environmental and Radiological Health Sciences Colorado State University, Fort Collins, CO 80523, USA Department of Kinesiology, Recreation, and Sport Studies University of Tennessee, Knoxville, TN 37996, USA Knoxville Regional Transportation Planning Organization City of Knoxville, TN, 37902, USA Corresponding author information: Jeremy D. Auerbach Department of Environmental and Radiological Health Sciences Colorado State University, Fort Collins, CO 80523, USA Phone: 970-491-1063 Fax: 970-491-2940 E-mail: [email protected]
Abstract
Introduction.
Student active commuting to school is an important component to student achievement and student health, yet this form of physical activity has significantly declined in the U.S. Distance between the school and student residence is often reported as a barrier for student walking, thereby increasing street and trail connectivity between and within residential developments and schools could foster student walking. The purpose of this study is to evaluate the potential benefits of increased thoroughfare connectivity on student walking within school walking zones.
Methods.
This study conducts a cost-benefit analysis of increased thoroughfare connectivity around elementary and middle schools in a U.S. school system that serves sixty thousand students. Benefits, which include the increased time of physical activity from student walking and the potential cost-savings to a school system if they had fewer students to bus to school, are compared to the financial costs of the new connections. Advanced network optimization techniques were applied to several suburban and rural schools from a representative the school system to locate the optimal new thoroughfare connections that maximize student walking to a school and minimize the length of the new thoroughfare.
Results.
Results from this case study showed that short and inexpensive new thoroughfares could increase the potential number of student active commuters and provide a significant increase of physical activity for those potential student walkers.
Conclusions.
This work can foster the integration of student walking and student health in residential planning decisions around schools.
1. Introduction
There has been a rapid decline in physical activity reported for U.S. children (Nader et al., 2008). This lack of physical activity is one of the primary contributors of childhood obesity and physical activity is a valuable component of childhood development and academic achievement regardless of socioeconomics, demographics, geography, and school characteristics (Robert Wood Johnson Foundation, 2009, Centers for Disease Control and Prevention, 2010). Increased physical activity has also been correlated with higher attendance rates and fewer disciplinary incidents (Welk, 2009).
One way to combat this trend of declining physical activity in children is to promote student active commuting to school, through walking or biking. Children who actively commute to school tend to be more active outside of this commuting to school and active commuting to school has been shown to be inversely associated with body mass index (Sirard and Slater, 2008, Lubans et al., 2011, Mendoza et al., 2011, Turrell et al., 2018). Not surprisingly, with the decline of physical activity in children in the U.S., there has also been a decline of active commuting to school (McDonald, 2007, Pedestrian and Bicycle Information Center, 2010, The National Center for Safe Routes to School, 2016). Parents have reported several behavioral and environmental barriers that have contributed to this decline: the perception of possible violence or crime along the route; the speed and volume of traffic along the route; poor weather or climate in the area; and the distance between home and school (Nelson et al., 2008, Pedestrian and Bicycle Information Center, 2010, McDonald and Aalborg, 2009, Centers for Disease Control and Prevention, 2005, Mendoza et al., 2014). In the U.S., centrally located school have been advocated to promote student walking since at least the 1920's (Perry, 1929). More centrally located schools have more students walking or biking, even after controlling for other neighborhood characteristics (Kim and Lee, 2016), but recent trends in the school site design and location in the U.S. make active transportation difficult (Chriqui et al., 2012). Since the 1970s, school systems have increasingly constructed larger schools on larger tracts of land, often in rural areas further from population centers with the number of schools dropping from 262,000 in 1930 to 95,000 in 2004, while the student population has increased from 28M to 54M (Office of Children's Health Protection, 2011). With these barriers and trends in the siting of U.S. schools not changing another avenue is to increase the walkability of the residential neighborhoods proximate to schools. This connectivity of the built environment has been found to be positively correlated with student active commuting as it contributes to a shorter distance between school and home (Frank et al., 2006, Babey et al., 2009, Bungum et al., 2009, Larsen et al., 2009, Giles-Corti et al., 2011, Coughenour et al., 2017, Marshall et al., 2014). Yet, the neighborhood design features that are currently popular with residents are at odds with thoroughfare connectivity. Residential neighborhood developers are incentivized to minimize thoroughfare connectivity in order to reduce the number of community entrances and increase the number of buildable lots, especially lots on cul-de-sacs. These built environment features are desirable for residents as they provide low-traffic streets and the perception of security associated with them. As expected, reduced neighborhood walkability and increased distance between home and school has been shown to be positively correlated with hildhood obesity (Spence et al., 2008, Grafova, 2008, Oreskovic et al., 2009, Tewahade et al., 2019). This decline in active commuting to schools also leads to increased traffic from private driving and school busing, which accounts for 10-14% of morning rush-hour traffic (McDonald et al., 2011), and increased pollution. The additional traffic intensifies parental safety concerns related to traffic and student active commuting. Still, the number of children killed and injured while walking or biking is dwarfed by the increasing rate of vehicle crashes, which are the leading cause of death among school age children in the U.S. (National Center for Statistics and Analysis, 2016). Furthermore, this decrease in active commuting also imposes a considerable economic burden on school systems with increased costs related to purchasing and maintaining buses, hiring drivers (an occupation with a high turnover rate), and fuel and insurance expenditures (DeNisco, 2015). A study of school administrators found that addressing the perceived safety concerns and increasing the number of sidewalks can increase active travel to schools (Price et al., 2011), and one the key elements of the National Center for Safe Routes to School is transportation planning approaches to ensure safe active commuting opportunities (The National Center for Safe Routes to School, 2016). These school siting trends and neighborhood design features contribute to urban sprawl and to greater distances between student homes and schools, and when coupled with a lack of sidewalks and bike paths greatly reduce student opportunities for active commuting (Kouri, 1999). The purpose of this study is to estimate the potential benefits of designing residential developments with more thoroughfare connectivity. In order to do that, it models how existing walking zones could be expanded via greater thoroughfare connectivity around schools. Several potential benefits may result from this research: (1) the study will examine the potential cost-savings to school systems if they had to provide busing for a smaller proportion of students, (2) the expected health impact to students (e.g., increased physical activity) by increasing the potential that they will walk or bike to school, and (3) an approachable methodology to automate the discovery of potential thoroughfares for planners, school officials, and researchers.
2. Methods 2.1 Case Study
We have worked closely with the Knox County School (KCS) public school system to evaluate the impact of increased thoroughfare connectivity on student walking. KCS serves 60,000 students in and surrounding Knoxville, Tennessee, with 89 schools and 337 buses. School acreage and attendance growth for schools in the KCS have undergone changes similar to those experienced by school systems across the US (Transportation Consultants, 2014). The average acreage of elementary schools jumped from 8.5 acres for schools built prior to 1977 to 24.5 acres for schools built since then (see Fig. 1). Due to school site selection in Knox County, from the 2013-14 school year to the 2017-18 school year, the number of students who lived outside the walking distance increased by approximately 6,000. As with most US school systems, KCS has an established policy to determine if a student is eligible for transportation by a bus, based on each student's residence location in relation to their school (Knox County Schools Transportation Department, 2009). A student is inside the school walk zone (SWZ), and ineligible for busing, if the distance from the tudent's residence to the drop-off location of the student's zoned school is less than 1 mile (for elementary students) or less than 1.5 miles (for secondary students), based upon the existing street network. Fig. 1. School acreage and attendance growth for schools in the Knox County School System. A report by the Knox County Department of Engineering and Public Works identified that 37.6% of the total number of students in the school district (22,322 students) lived within the SWZs (Transportation Consultants, 2014). They calculated the shortest path distance for these students, based on the thoroughfare topology, and found more than 30% of those walking distances were longer than the SWZ distance. Therefore, KCS is responsible for busing large numbers of students (up to 7,434 students) to school who could potentially be within active transportation distance, which places a significant financial burden on the school district. A significant contributor to this to distance from residence to school is poorly connected street networks. An analysis conducted by the authors of this study found many Knox County schools were near neighborhoods with street networks that had few intersections or frequent dead-ends (cul-de-sacs), i.e. neighborhoods with low thoroughfare connectivity. As such, the network SWZs fail to capture many students that reside close to the school βas the crow flies," that is, the straight-line or βEuclidean distance" (see Fig. 2.A). Fig. 2. An example of the benefit of a new thoroughfare connection. In figure A orange nodes represent the residential parcels within the Euclidean SWZ distance (blue dashed line) of the school (blue square) but not within the network SWZ distance (black lines denote streets). In figure B, the residential parcels within the network SWZ distance are represented by green nodes and the residences outside of the Euclidean SWZ distance are the red nodes.
Since student residences for a given school can change over the course of the year and from year to year, as students enroll, graduate, or move, we used residential parcels as the proxy for student residence locations. Data about residential parcel locations and types, such as single-family residence or multi-family residence, were provided by GIS administrator for Knoxville, Knox County, and the Knoxville Utilities Board. Knoxville-Knox County Planning has provided the average number of students for residential parcel types in the study area, which was used to estimate the number of students that would be affected by changes in thoroughfare connectivity, the number of students for each school in Knox County, the number of students within each school's SWZ, and the number of students that are within the SWZ and network distance to their respective school. The data was collected in 2017 and engineered according to the methods provided in the Supplemental Materials.
Ten schools that would benefit the most from additional thoroughfare connectivity were selected for analysis with the SWZ distance disparity metric. SWZ distance disparity was developed for this study and is the difference between the number of students in the Euclidean SWZ distance and the number of students in the network SWZ. A large proportion of students within both the Euclidean SWZ distance and the network SWZ is associated with an environment that has high thoroughfare connectivity, typically urban. A large proportion of students within the Euclidean SWZ distance and a low proportion in the network SWZ is indicative of a built-up yet low-connectivity environment, typically suburban. Low proportions in both the Euclidean SWZ distance and the actual SWZ is associated with a low-density, typically rural, environment. The ten schools with the largest SWZ distance disparity were selected for the analysis in this study (see Fig. 3.). These ten schools were from suburban and rural environments, as urban schools already ad large proportions of students within the network SWZs and therefore low SWZ distance disparity. Fig. 3. SWZ distance disparity for Knox County schools. The ten schools with highest SWZ distance disparity, i.e. the number of students who would be in the SWZ if there were additional street and trail connections, are labeled and selected for analysis.
Combinatorial optimization techniques were employed to identify and evaluate new street connections (see Fig. 2.), expanding on sidewalk siting analysis (Randall and Baetz, 2001). Fig. 2. A provides an example of residential parcels (orange nodes) within the Euclidean SWZ distance (blue dashed line) of the school (blue square) but not within the network SWZ distance. The residential parcels within the network SWZ distance are represented by green nodes and the residences outside of the Euclidean SWZ distance are the red nodes. In Fig. 2. B the dashed line depicts the optimal new connection that maximizes the number of residences now within the network SWZ distance and minimizes the length of the connection. To find the optimal thoroughfare that maximizes the number of residences within the network SWZ, exhaustive search optimization algorithms were performed in MATLAB (version 9.3 R2017b, see Algorithm 1). The exhaustive search optimization routine creates an edge for combinations of nodes and if that shortest path distance is less than the SWZ distance then it evaluates the cost of the new connection, i.e. the length of the new connection, and the benefit of the new connection, i.e. the number of residences that are now within the network SWZ distance (see the Supplemental Materials for the algorithm pseudocode). Note we parallelize the optimization routine as finding all of the solutions or large networks is computationally expensive (thousands of nodes results in millions of possible connections to evaluate). Heuristics have been developed to find near-optimal solutions when the exhaustive search is not computationally feasible (Auerbach,2018). For the optimization algorithm, the network distance between two nodes is given by d(i, j) , and if the distance between a node and the school node ( S ) is less than the SWZ distance, d(i, S) < D (where D is 1 mile for elementary schools and 1.5 miles for middle schools) then the node i is assigned to the set of close nodes ( N C ), otherwise it is assigned to the set of distant nodes ( N D ). Connections are made between two nodes, one being from N C and no restrictions on the other node. After a new connection is established, the residential nodes that are now within the network SWZ distance are assigned to the set N C . The cost of the new connection is the connection length, C(i, j) = d(i,j) , and the benefit of this new connection
B(i,j) is the number of additional residences now within the SWZ. The optimal solution is the solution with the greatest benefit, or number of new residences now within the distance to the school, which can be expressed as the bi-objective function π β = max (π,π) (π΅(π, π) β πΆ(π, π)). The cost of a new connection was based on the average national cost for local or state governments to build a sidewalk (5 ft wide), approximately $1 million per mile, excluding right-of-way, crossing water/wetlands, and major topography (Bushell et al., 2013). The national average busing cost for a student was estimated to be approximately $1,000 in 2017 (using the most recent data from 2012-13 and adjusted for inflation) and used for the economic benefit of a new connection (National Center for Education Statistics, 2016). To estimate the health benefit of potential new street and trail connections, the number of additional residences within the SWZ were converted to the number of potential students who could walk or bike to school. The total benefit was calculated by taking this number of students and multiplying it by the estimated duration of the physical activity. The average U.S. student who walks to school acquires 16 minutes of moderate-intensity activity and they walk at a rate of 71.7 meters per minute (Crouter et al., 2013). Assuming a linear relationship between distance walked and walking time, it would take an average student 22.5 minutes to walk 1 mile, and this rate was used to calculate the walking benefit for the students along the optimal connection for each school evaluated.
3. Results
The location of the optimal connections are given in Fig. 4. The black square represents the school, and grey lines indicate streets. Green nodes are residences within the network SWZ distance, black nodes signify residences not within the network SWZ distance and orange nodes are the residences that are within the Euclidean SWZ distance after the new connection is made, denoted by the orange line. The length of the optimal connection, the number of residences and students included in the network SWZ with the new connection, and the total time spent walking to and from the school for these students are provided in Table 1. For each school the optimal student walking time was calculated by randomly selecting a number residences by the proportion of students per esidence for the given school. A thousand random samples were used to calculate the average and standard deviation for the optimal walking time for each school. Fig. 4. Optimal connection results for the ten schools. The black square represents the school, and grey lines indicate streets. Green nodes are residences within the network SWZ distance, black nodes signify residences not within the network SWZ distance and orange nodes are the residences that are within the Euclidean SWZ distance after the new connection is made, denoted by the orange line. Table 1. Active commuting characteristics and optimal results for the ten schools with the greatest walking potential. Schools denoted with an * are located in a suburban area, otherwise the school is located in a rural area. Standard deviations for the optimal walking times are given in the parentheses.
Proportion of students within walking
Elementary
A* 0.11 0.28 292 1680 0.17 453 3704 77 2936 (30) B 0.01 0.03 239 1035 0.23 137 3891 39 1308 (16) C 0.05 0.18 228 1173 0.19 203 3613 39 1536 (24) D* 0.41 0.53 220 1911 0.12 255 3133 31 1168 (26) E 0.06 0.19 210 1154 0.18 287 3307 52 2008 (30) F 0.08 0.35 173 1120 0.15 287 2907 43 1502 (32) G* 0.20 0.31 171 1027 0.17 703 602 120 4220 (76) Middle
H* 0.10 0.29 316 2033 0.13 940 3260 122 6452 (112) I 0.04 0.17 269 1058 0.25 464 3554 116 5994 (100) J* 0.29 0.45 209 4246 0.05 381 5027 19 1104 (28)
4. Discussion ccording to the results, increased connectivity could save the school system up to $120K/year in reduced busing costs for a one-time sidewalk cost as little as $114K, per school. These additional connections would also provide an elementary school students an additional 40 minutes, on average, of moderate-intensity physical activity (MVPA) from walking each school day (approximately 50 MVPA minutes gained for a middle school student) (Bassett et al., 2013). This is a significant amount of physical activity that could meet two-thirds of the national recommendations for physical activity (i.e., 60 minutes or more of MVPA per day) and help foster activity outside of commuting, reduce delinquencies, increase academic achievement, and reduce student BMI (U.S. Department of Health and Human Services, 2018). Other additional benefits to increased active commuting to school, include reduced traffic, and its associated reduction in air pollution and accidents, and reduced healthcare costs. According to 2016 data, in the U.S. 1 in 5 school-age children are obese (Hales et al., 2017) and obese children ages 2-19 in the United States were found to stay on average 0.85 days longer for hospital treatment, incur $2000 additional charges, and $925 more in hospital costs than non-obese children (in 2017 dollars) (Trasande et al., 2009). Furthermore, obese children incurred $210 in higher outpatient visit expenditures, $124 higher prescription drug expenditures, and $13 higher emergency room expenditures (Trasande and Chatterjee, 2012). There were approximately 73.8 million children under the age of 18 in the United States in 2017 (Flood et al., 2018) and assuming obese children were hospitalized at the same rate as non-obese children (children were hospitalized in the U.S. at a rate of 0.0014 in 2012 (Witt et al., 2014)), this translates to an estimated $40 million dollars in additional hospital costs and $20 million in additional charges for the hospital treatments of obese children in 2017. There are barriers to including walkabaility in residential development as Knox County, TN, like many other places in the U.S., has one group that make decisions about school siting and busing policies (the local school board) and another that makes decisions about the design of new neighborhoods and other developments (the local planning commission). Those groups may not always be aware of how decisions they make affect each other. These results describe and quantify the ways in which decisions made by the planning commission, specifically about thoroughfare connectivity in developments near schools, can impose additional costs on school systems, in the form of the need to provide more busing. Reduced thoroughfare connectivity around schools also imposes costs on families and on children themselves, in the form of less opportunity for physical activity, potentially lower academic achievement, and increased risk of obesity and other health problems. It would be in the best interest of school systems to request that the planning commissions take into account the impact that low thoroughfare connectivity around schools has on busing costs. It is probably too late to remedy the low thoroughfare connectivity around the schools examined in this paper, because of cost and likely neighborhood opposition. But the key finding of this paper -- that short connections can vastly increase connectivity and walkability around schools, while decreasing future busing and health costs -- should inform future decision-making about neighborhood design around schools with the goal of reducing costs and improving students' health and academic achievement. An expanded working relationship between planning commissions and school systems may eventually lead to additional opportunities to collaborate toward cost savings, such as better coordination of school siting with other land use goals. The methods provided here are an accessible approach to evaluate walkability for regional planners and local school officials. Results from this work have been shared with Knox County planning commissioners, the Knox County School Board, and other decision-makers with the intent of making them aware of the costs imposed by lack of thoroughfare connectivity in neighborhoods (Thaler and Sunstein, 2008). They could also influence school officials to consider thoroughfare connectivity and student active commuting when siting schools. Furthermore, these results could also start the dialogue of allowing greenways to be included as student commuting paths for school systems that currently do not consider them as permitted thoroughfares for student active commuting. The presence of greenways near residences has been shown to increase property values (Nichols and Crompton, 2005) as prospective home buyers are willing to pay more for a home in a walkable neighborhood (Knoxville Area Association of Realtors, 2017). After optimal connections are computationally identified, residents should be involved with the design of their future communities (Forester, 1999). Finally, school accessibility should be included in any research using an ecological systems approach or considering the built environment as a factor in childhood obesity (Fig. 5. modified from Davison and Birch, 2001). Fig. 5. Ecological systems theory approach to childhood obesity modified to include the accessibility of school (modified from Davison and Birch, 2001).
5. Limitations
There are several issues with the current analysis. The study area is limited to one U.S. county, and to make these results more robust, the networks for schools in additional study areas should be evaluated. There are several methodological issues. Residences may be placed on the nearest street but that may not be the street they are physically located on. The error rate for this occurring is small for this study area but could vary for different study areas. Restrictions on the locations of ew streets and trails was not included. For example, thoroughfares cannot be built on slopes beyond a specific threshold, and it may be cost prohibitive or impossible to route them through specific parcel types (such as commercial areas, mining sites and landfills) or highways. Solutions are dependent on the location of nodes at every parcel and intersection. This can lead to large sparse areas of possible connectivity and could be corrected with the placement of artificial nodes at regular intervals at the expense of computational costs. The issue that newly constructed connections can intersect existing streets (planarity) was not resolved in this study. Calculating the determinants for each pair of lines is computationally expensive. Lastly and due to the limitations of shapefiles, it is recommended that government agencies move away from their dependence on them for spatial data. There are several emerging data formats, such as OGC GeoPackage and GeoJSON, that are robust and free.
6. Conclusions
The results show that even the addition of short thoroughfares can significantly increase the number of residences within the walking zone of a school. These additional residences now within the walking zone increase the potential number of students who could actively commute to school, and commuting is positively correlated with improved health and education achievement. Planners and other decision-makers can use this result to increase their awareness about the costs, to families and to school systems, imposed by the lack of thoroughfare connectivity in new neighborhoods. We offer an approach to evaluate school walkability for regional planners, school systems, and health departments.
Acknowledgments
We would like to thank Alex Zendel (GIS Analyst at the Knoxville-Knox County Metropolitan Planning Commission) for compiling the school network data, and Rickey Grubb (Director of Enrollment and Transportation at Knox County Schools) for relevant information about school busing costs.
Contributions:
Jeremy Auerbach:
Conceptualization, Methodology, Software, Validation, Formal Analysis, Investigation, Data Curations, Writing - Original Draft, Writing β Review & Editing, Visualization
Eugene Fitzhugh:
Conceptualization, Writing - Original Draft, Writing β Review & Editing, Funding Acquisition
Ellen Zavisca:
Conceptualization, Writing - Original Draft, Writing β Review & Editing
Funding:
JDA and ECF were supported by a University of Tennessee Transportation Center Fellowship award.
References upplemental Materials
Algorithm Psuedocode
Exhaustive search pseudocode for i in N D do Select a distant node for j in N C do Select a close node if d(i,j) + d(j,S) < D then If the new connection changes the distant node to close
C(i,j) = d(i,j)
Calculate the distance of the new connection for k in N D do Select a distant residence if d(k,i) + d(i,j) + d(j,S) < D then Calculate the distant to the school k π π πΆβ² If the distant residence is within the SWZ, then label it close end if end for
B(i,j) = | π πΆβ² | Calculate the number of new close residences end if end for end for
Data Engineering
The spatial data provided by KGIS was in shapefiles, which are commonly used by government agencies but have significant limitations for analysis. The data cleaning and processing methods described below are provided for planners and researchers to use for their own systems (see Karduni et al., 2016, and Oliver et al., 2007, for additional useful methods to process shapefile data for network analysis). To identify optimal thoroughfares a network approach was used and the data was converted to a network with nodes at residences and street intersections and the streets were converted to edges. To accomplish this data transformation, in ArcMap (version 10.6): 1. buffers were created around each school (1 mile for elementary schools and 1.5 miles for middle schools). 2. street edge and street intersection node development: a. street data outside of the school buffer were removed (using the clip tool), b. highways were removed, as students will not use them for active commuting, c. streets part of the school property were also removed (as schools do not want direct connections), d. nodes at the street intersections were created and labeled as intersections, i. points at the street intersections were created (intersect tool with output type point), ii. two numeric fields were added to this layer using a field type DOUBLE to ensure maximum precision and the spatial coordinates were identified for the street intersection points (calculate geometry), iii.