Leveraging Human Mobility Data for Nonmotorized Traffic Monitoring: A Solution for Sparse Data

Conceptually, human mobility data (that are collected from mobile devices) can help generate adjustment factors across the entire road network with data passively collected on a continuous basis covering a large geographic region. This study explores whether and how human mobility data could be used in generating adjustment factors for nonmotorized traffic (i.e., walking and biking activity) monitoring purposes. This study first found that adjustment factors calculated with link-level Strava Metro data are at least within a similar range as what can be obtained from permanent counters. Then it was observed that adjustment factors calculated based on the two different data sources (i.e., Strava Metro and permanent counters) are of medium to high correlations in about 40% of the cases. Statistical models were then estimated to look for significant variables from various sources that could potentially help improve the usefulness of the adjustment factors calculated with Strava Metro data (i.e., calibration). The two models estimated separately for walking and biking activities suggest that sociodemographics (related to age), land use variables (i.e., site count by place category), and human activity variables (i.e., visit frequency by place category) are significant explanatory variables. In the validation case, adjustment factors calculated with Strava Metro data after applying parameters from the estimated models for calibration outperform the factors without any calibration in predicting short-term counts. The proposed methodology could assist areas with limited permanent counters to advance their practice in short-term count extrapolations, network-wide nonmotorized traffic predictions, and counter location selections.