Varying influences of the built environment on daily and hourly pedestrian crossing volumes at signalized intersections estimated from traffic signal controller event data

Direct-demand models of pedestrian volumes (identifying relationships with built environment characteristics) require pedestrian data, typically from short-duration manual counts at a limited number of locations. We overcome these limitations using a novel source of pedestrian data: estimated pedestrian crossing volumes based on push-button event data recorded in traffic signal controller logs. These continuous data allow us to study more sites (1494 signalized intersections throughout Utah, US) over a much longer time period (one year) than in previous models, including the ability to detect variations across days-of-week and times-of-day. Specifically, we develop direct demand (log-linear regression) models that represent relationships between built environment variables (calculated at ¼- and ½-mile network buffers) and annual average daily and hourly pedestrian metrics. We control spatial autocorrelation through the use of spatial error models. All results confirm theorized relationships: There is more pedestrian activity at intersections with greater population and employment densities, a larger proportion of commercial and residential land uses, more connected street networks, more nearby services and amenities, and in lower-income neighborhoods with larger households. Notably, we also find relevant day-of-week and time-of-day differences. For example, schools attract pedestrian activity, but only on weekdays during daytime hours, and the coefficient for places of worship is higher in the weekend model. K-fold cross-validation results show the predictive power of our models. Results demonstrate the value of these novel pedestrian signal data for planning purposes and offer support for built environment interventions and land use policies to encourage walkable communities.