Background
Studies examining built environment influences on walking and physical activity often incorporate objective measures of street connectivity as predictors. Little research has been conducted to study the influence of modeling off-street pedestrian paths (Chin et al. 2008) on travel choice. Off-street pedestrian paths such as cul-de-sac connectors are likely particularly important for populations reliant on active modes of transportation such as youth without cars.

Objectives
The current study addresses two main objectives, drawing on data from the Region of Waterloo, Ontario. First, we quantify how off-street pedestrian path connectivity can improve pedestrian access. Second, we highlight the implications of measuring these paths for walkability research, by comparing how alternate connectivity measures explain walking trips for samples of adults (n=1,447) and youth (n=1,100).

Methods
This study draws on objective data from the NEWPATH (Neighourhood Environment in Waterloo Region: Patterns of Transportation and Health) Project, a transdisciplinary research program with partners from the University of British Columbia, University of Alberta, University of Waterloo, and the Region of Waterloo. A pedestrian network for the Waterloo Region was created from three components using Geographic Information Systems: the local street network, cul-de-sac connectors, and multi-use trails. Two measures were derived to assess pedestrian access: buffer area and intersection density. Sensitivity analyses were performed to assess differences in these measures and their influence on walking behavior, based on the network used to generate the measures: the entire pedestrian network or a network based on specific sub-components, e.g. the local street network.

Results
Results indicate that access, as gauged by buffer area, is highly sensitive to the inclusion of off-street pedestrian linkages. Inclusion of these linkages results in an average increase in buffer area of 30%. Both cul-de-sac connectors and multi-use trails contribute similarly to this increase, but the distribution of cul-de-sac connectors exhibits more spatial variation, translating into more pronounced increases in pedestrian access in low walkability areas. These areas are characterized by large numbers of cul-de-sacs and the inclusion of cul-de-sac connectors therefore results in buffer size increases of greater than 40%. In addition, 75% of cul-de-sacs are disconnected, highlighting the potential for improved pedestrian access through the incorporation of more connectors in future developments.

When calculated using the pedestrian network, intersection density is estimated at 23% higher then when calculated based solely on the street network. This increase varies modestly by walkability, from 19% in low walkability areas to 24% in higher walkability areas. To isolate the effect of off-street pedestrian connections on walking behavior, two logistic regression models were created, one with a street-network based intersection density measure as a predictor of walking at least once over a two day period, and one with a pedestrian network based intersection density measure as a predictor. In both models, age, gender, household income, household size and car ownership were entered as control variables. Although intersection density calculated using the pedestrian network should in theory better predict walking than a comparable measure calculated using solely the street network, this was not found to be the case for the sample of adults (n=1,447). In both models, intersection density was found to be a highly significant predictor of walking, but the odds ratio for the street network based measure was 26% higher than that for the pedestrian network based measure. Similar analyses are currently being conducted for a sample of youth. It is hypothesized that the incorporation of off-street pedestrian paths in a measure of intersection density will better predict walking behavior for youth as a population more reliant on active modes for transportation.

Conclusions
Results indicate that the failure to include off-street pedestrian paths when modeling pedestrian access may result in substantial underestimation of the area accessible to pedestrians. Further, this discrepancy varies by walkability, with greater discrepancies in less walkable areas corresponding to higher levels of off-street pedestrian infrastructure. The results also highlight considerable potential for off-street pedestrian infrastructure improvements in the Region of Waterloo. Despite these findings, a measure of intersection density based on a pedestrian network comprised of both street and off-street pedestrian linkages was not found to better predict walking than a comparable measure based solely on the street network, suggesting the need for future research in this area.

Reference
Gary K.W. Chin, Kimberly P. Van Niel, Billie Giles-Corti, Mathew Knuiman. 2008. Accessibility and connectivity in physical activity studies: The impact of missing pedestrian data, Preventive Medicine 46(1), January pp 41-45.

Support/Funding
Heart and Stroke Foundation and Canadian Institute for Health Research.