Overall agreement

Table 1 displays overall levels of agreement between all data sources and for each individual data source. 50.1% of the total area enclosed by any data source’s park boundaries is enclosed by all data sources, while 33.0% of the total enclosed area is represented only in one data source. The PPR boundaries have the lowest total area (39.86 km) and the greatest level of agreement with the other data sources, with 88.8% of the PPR’s area enclosed by all other data sources. OSM has the highest total area and the lowest level of agreement, with 30% of its total area not being shared by any other data source’s boundaries. This result is also displayed in Map 1.

Table 1 Rows display the number of data sources providing overlapping boundaries. “No overlap” indicates areas that are enclosed by boundaries in only one data source, while “complete agreement” indicates areas that are enclosed by the boundaries of all four data sources. 2-way agreement and 3-way agreement indicates areas that are enclosed by any two or any three data sources, respectively. Data-specific columns show agreement between each individual data source and the rest of the groupMap. 1

Overall agreement between data sources. Map. 5 symbolizes the overall agerement between data sources by showing the overlap of park and greenspace boundaries across the city of Philadelphia. Darker features are represented in more data sources than are lighter features, with the darkest color symbolizing overlap of all data sources and the lightest color symbolizing that only one data source includes that feature

Comparison of parkland/greenspace by planning district

For each of 18 planning districts in Philadelphia, Table 2 displays the proportion of park/greenspace as calculated from each source of assessed data, as well as the overall mean and coefficient of variation. The table shows reasonable agreement between the data sources at the district level for most districts. For some districts, such as the Lower Far Northeast, University Southwest, and Lower Southwest, there are large discrepancies. Depending on the neighborhood of residence for a study participant, different data sources could provide considerably different area-level exposure estimates for parks or greenspace. This result is also displayed in Map 2.

Table 2 Proportion of parks or greenspace in each planning district by data sourceInclusion of informal greenspaces

A total of 9,663 greened vacant lots were identified as part of the LandCare Program. Of these, 8 lots were within a PPR property; 1,914 within an OSM park or greenspace; 7 within a DVRPC protected open space; and 15 within a PAD-US-AR greenspace. At points of intersection, these maps tended to reflect the same LandCare lots: 7 lots had their center in all four sources of data.

Overall, a small proportion of greened vacant lots were included in any of the park data, with Open Street Maps including the largest number. Despite being an important source of greenspace for Philadelphia residents, greened vacant lots are not represented except in specialized maps.

Qualitative analysis

Visual comparisons of the maps in selected areas of Philadelphia provided insights into the similarities and differences in how each data source addressed jurisdictional boundaries, intersecting roads or highways, and buildings. We also evaluated the mapping of the Schuylkill River Trail with attention to the completeness of its representation.

Jurisdictional concerns

Philadelphia parks and greenspace are managed by a diverse group of organizations, including national, state, and local parks as well as private and informal parks. Organizations publishing maps of parks and greenspace in Philadelphia may limit their maps to certain kinds of parks. With the exception of the PPR Properties map, the data assessed here include municipal, state, and national parks (such as the John Heinz National Wildlife Refuge) but all maps vary in their inclusion of private parks. Both OSM and PAD-US-AR include private greenspace managed by the University of Pennsylvania and Drexel University in the University City area of Philadelphia, but the exact boundaries vary, and the DVRPC and PPR maps exclude them entirely. Only OSM includes graveyards and cemeteries, some of which have initiatives to increase outdoor recreational use [38, 39].

Inclusion of road and highway areas

Each data source included highway and road crossings in different ways. As illustrated in Maps 3, 4, 5, local roads that bisect parks are included within park boundaries for all data sources. However, intersections with interstate highways are represented differently. Fairmount Park is a large park that covers both the east and west sides of the Schuylkill River. For example, Interstate 76, a major north-south federal controlled-access highway, runs parallel to the river, with a small buffer between the highway and the river that incorporates a local roadway, a mixed-use trail with points of river access, and some riparian vegetation. Map 3 illustrates that only the PPR map excludes the majority of Interstate 76 from Fairmount Park, though it does include accessible pedestrian underpasses (Fig. 1). This means that GPS traces following roadways that intersect large parks, for example from Interstate highway drivers, could be categorized as park visitation.

Map. 2

Philadephia planning districts. Panels display Philadelphia planning districts, symbolized by the proportion of area of each district that is categorized as greenspace or parkland by four different sources of data: the Philadelphia Parks and Recreation Department, Open Street Maps, PAD-US-AR, and the Delaware Valley Regional Planning Commission. Individual districts may fall into different categories of park/greenspace density depending on the source of data used, because different data sources delineate different boundaries and have different eligible properties. For example, the Lower southwest contains the John Heinz National Wildlife Refuge, which is included in all sources of data except PPR

Map. 3

Fairmount Park. Panels display the area surrounding Fairmount Park, comparing map data from the Philadelphia Parks and Recreation Department, Open Street Maps, PAD-US-AR, and the Delaware Vally Regional Planning Commission. While all maps display similar boundaries for Fairmount Park itself, smaller park areas within the larger park are treated differently. While PPR and DVRPC exclude the Schuylkill River from the boundaries of the park (1), Open Street Maps and PAD-US-AR includes the river. Only PPR excludes Interstate 76 from the park features (2), including as parkland only pedestrian-accessible underpasses. Only PAD-US-AR excludes rain infrastructure from the park features (3). No data excludes local roads from park features, although this area includes large and heavily-trafficked roadways such as Belmont and Montgomery Avenues to the west and Kelley Drive to the east. All maps other than OSM include the Philadelphia Zoo (4)

Map. 4

Pennypack Park and surroundings. Panels display the area surrounding Pennypack Philadelphia, comparing map data from the Philadelphia Parks and Recreation Department, Open Street Maps, PAD-US-AR, and the Delaware Vally Regional Planning Commission. While all maps display similar boundaries for Pennypack Park itself, Open Street Maps includes fields between runways of the Northeast Philadelphia Airport as grassland (1), and PAD-US-AR operationalizes the median of an urban highway (2). PAD-US-AR is also the only data source that includes the entirety of the Abraham Lincoln High School campus as a park (3). PPR is the only data source to include the Lorimer Trail Greenway (4)

Fig. 1

Sweetbrair Drive provides a pedestrian-accessible path underneath I-76 in Fairmount Park and is included in the PPR map

Similarly, Map 3 illustrates that PAD-US-AR categorizes the entirety of a surface-level federal highway (Roosevelt Boulevard/US Route 1) as parkland. This Boulevard has grassy or tree-lined medians for much of its length that are accessible via pedestrian crossings. However, in other areas, the median consists of only a narrow strip of grass that is made largely inaccessible by heavy motor vehicle traffic and limited pedestrian crossings (Fig. 2), and the PAD-US-AR map includes both the medians and the roadway itself within the park area.

Fig. 2

Roosevelt Boulevard is included in the PAD-US-AR map

Another major road in Philadelphia, the Benjamin Franklin Parkway, is included differentially between data sources. The Parkway includes medians with sidewalks and tree canopy, as well as enclosing the multi-use public open space Eakins Oval. In addition, the Parkway is periodically closed to motor vehicle traffic to accommodate local events such as bike rides and running races [40]. At different times, both the roadway and its medians can function effectively as public parkland. Both OSM and PAD-US-AR include the entirety of the Parkway as a park feature, while PPR and DVRPC include only the medians and Eakins Oval.

GPS traces that follow roads or highways through a park could be classified as either exposed or not exposed depending on the data source and the road in question.

Inclusion of buildings

While publicly-accessible buildings can have functions that are similar to parks, such as facilitating social contacts, education, recreation, and exercise, researchers may wish to isolate exposure to outdoor park spaces. Park boundaries include building footprints in all data sources. The boundaries of Fairmount Park in Map 3incorporate multiple buildings, including a children’s museum and several historic buildings and homes. Independence Mall is a national historical park with multiple buildings and museums; these are included in park boundaries among all mapped data except PPR, which does not manage this site.

Schuylkill River Trail (SRT)

Urban greenways are often important greenspaces for cities because they can support recreational uses as well as walking and cycling for transportation [41]. None of the data sources fully cover the SRT. Each map excludes the length of the SRT north of Shawmont Avenue. Further, the maps differ at intersections. Map 5illustrates the location where the SRT passes underneath a major highway interchange at City Avenue. Each data source chooses a different pattern of exclusion and inclusion. While users of the SRT pass safely underneath the highway overpass, separated from road traffic in the vertical dimension, this proves difficult to fully map in two dimensions. PAD-US-AR inappropriately excludes all intersecting highway areas and OSM inappropriately includes all highway areas even where there is no park or greenspace underneath (See Map 5).

Map. 5

Schuylkill River Trail. The Schuylkill River Trail is a linear park that forms a substantial component of Philadelphia’s park infrastructure, and that in some places proves difficult to represent in mapped data. These panels display the area where the Schuylkill River Trail, alone the northeastern bank of the Schuylkill River, passes underneath US-1. In data from the Philadelphia Parks and Recreation Department and the Delaware Valley Regional Planning commission, this is represented as a thin feature along the riverbank that intersects the highway (1). In contract, PAD-US-AR excludes the entire highway interchange (2), despite the fact that there is parkland underneath it, and Open Street Maps includes the entire area as a continuous feature (3), although some areas are inaccessible due to the highway infrastructure