Walking is the most common physical activity [1]. People who walk more in their leisure-time or as a form of active transportation have lower risks for developing type 2 diabetes, cardiovascular disease, certain cancers, and premature mortality [2–4]. Active transportation also reduces automobile use and greenhouse gas emissions, thereby improving air quality and decreasing acute and chronic respiratory illnesses within the population [5, 6].

In an effort to increase the physical activity and active transportation rates in the population, researchers have been studying the determinants of these behaviors. One such determinant is the walkability of neighborhoods and communities [5, 7–11]. Factors affecting walkability include the speed limits and connectivity of roads and the presence of sidewalks. Roads with high speed limits and roads that are poorly connected can make it unsafe and inefficient (i.e., long travel distances) for people to walk on and use for active transportation [7–9]. Sidewalks provide a pedestrian right-of-way on the roadside, and not surprisingly, traffic accidents involving pedestrians are far less common on roads that contain sidewalks [12, 13]. Furthermore, the distance of sidewalks within a neighborhood, as well as the proportion of roads with sidewalk coverage, predicts active transportation. For instance, the active transportation to school literature has demonstrates that walking to school is positively correlated with the sidewalk coverage along roads (e.g., percentage of roads in the school neighborhood with sidewalks) [14, 15]. Furthermore, the construction of new sidewalks along existing roadways contributed to increased active transportation to school rates within several communities that participated in California's safe routes to school intervention [16].

The walkability features of roads are typically measured using geographic information systems (GIS). GIS data on roads is widely available in most, if not all, industrialized countries. For instance, the CanMap^® RouteLogistics database (DMTI Spatial Inc., Markham, ON) provides geospatial information on public roads across Canada. GIS can also be used to measure the length and connectivity of sidewalks; however, GIS sidewalk data are far less common than road data and can be difficult and in some cases impossible to access. In Canada, for example, GIS sidewalk data are only available in larger cities, and this information is difficult for researchers to access as many cities do not share their GIS data. Furthermore, the GIS sidewalk data are collected and maintained in a different manner in different cities. Collectively, these issues make it challenging to conduct sidewalk-based research studies in smaller municipalities and in larger geographical areas (e.g., state/province or national level).

The purpose of this technical advance was to present a new GIS method for measuring sidewalk distances that can be used across regions and even entire countries. The method relies on an existing GIS road database and freely accessible Google Earth^® mapping software. We also tested this method in 315 locations across Canada to illustrate how it can be used. Although this method was developed and tested in Canada using a Canadian road network database, it could be easily extracted to other countries and databases.

The purpose of this technical advance was to present the protocol of a new GIS method for measuring the distance and coverage of sidewalks along roadways. The new sidewalk measurement protocol could be used in a variety of research, urban planning, and public health settings. The protocol could be used in descriptive research studies where sidewalk characteristics are being compared across different neighborhoods, communities, etc. It could be used in etiological research studies that aim to characterize the relationship between the presence and length of sidewalks with walking behaviors. Municipalities that do not have geospatial sidewalk data could use this protocol to create such information. Public health and educational authorities could use the sidewalk measures obtained in this protocol, in combination with road data, to identify safe routes for school-aged children to use when walking to school (i.e., all road segments used on travel route must have a sidewalk).

While obtaining the sidewalk measures in our testing sites we encounter three common issues. We discuss those issues here as well as the solutions we used to overcome them. First, for private sections of a neighborhood and gated communities images were not always available at the street level. In these instances, we were required to zoom out to the aerial view level. Second, some of the roads that were ≥ 4 lanes wide only had street view images for one side of the road. In these instances we navigated to the nearest intersection and did a 360 degree panoramic view to see if a view of the other side of the road was available, and if it was not, we zoomed out to the aerial view level. Third, in densely populated urban cores there were often trucks, buses, construction sites, etc. that blocked the view of the side of the road. In these instances we navigated forward or backward along the road at the street view level until we could see around the object(s), or if need be, zoomed out to the aerial view level.

The sidewalk measurement protocol developed in this study could be expanded in a number of ways. First, in addition to measuring the distance of sidewalks, it would be possible to measure the connectivity of sidewalks, in a similar manner to how road (street) connectivity measures are obtained. In fact, because 48% of the roads in the test sites of this national study did not have sidewalk coverage, it may be more appropriate to measure sidewalk connectivity rather than road connectivity in active transportation focused studies. The sidewalk connectivity measures could rely on the same type of indicators that are used to measure road connectivity such as the density of intersections per unit area, the percentage of intersections that are 3- or 4-way intersections, and the number of dead ends [7]. Researchers could take the sidewalk shapefiles saved at the end of Stage 2 in the protocol (i.e., road network that contains sidewalks) and obtain the connectivity measures on these files instead of the complete road network.

It may also be possible to integrate qualitative measurements of the sidewalks into the protocol, such as the sidewalk surface (e.g., pavement, asphalt, brick, etc.), sidewalk condition (e.g., excellent, in need of some repairs, in need of major repairs), and proximity of the sidewalks to the roads (e.g., directly beside road, boulevard separating road and sidewalk). Thus, rather than measuring the distance of all sidewalks, these distance measures could be broken down by qualitative features (e.g., distance of sidewalks in poor, fair, and good condition). Previous built environment research on parks has demonstrated that Google Earth can be used to obtain valid qualitative data [17]. It would be possible to integrate this type of information into Stage 2 of the sidewalk measurement protocol. More specifically, each of the road segments with a sidewalk could be coded based on its qualitative features (i.e., rather than deleting a road segment; code it a different color based on its qualitative features). Thus, in Stage 3 the distances of sidewalks that meet different characteristics could be calculated.

A key advantage of the newly developed method is having the ability to measure sidewalk distances and coverage in a consistent manner in several different municipalities. This will allow researchers who are interested in these types of studies to conduct regional, national, and even international studies. A key limitation of the newly developed method is that Google Earth does not provide street view and high resolution aerial images in all areas. In particular, these types of images are not available in many rural and remote areas. Another main limitation is the timing of when the Google Earth aerial and street view images were obtained relative to when the research study is completed. For example, in our city (Kingston, Ontario) the aerial images for Google Earth were obtained in 2004 and the street view images were obtained in 2009. If we were to conduct a Kingston-based study in 2012, some of the sidewalks in the city would have changed since 2004 and 2009, particularly in newly developed areas. Thus, as with most GIS measures of the built environment, the sidewalk measurement protocol would have a limited utility in newly developed areas.

The measurement of built environment constructs is becoming an increasingly important component of public health research. This study provides a new measurement protocol that researchers can use to measure the distance and coverage of sidewalks along roadways. It is hoped that the use of this protocol in future studies will lead to an improved understanding of the walking environment and the determinants of walking in different areas.