Planners and decision makers have increasingly voiced a need for network-wide estimates of bicycling activity. Such volume estimates have for decades informed motorized planning and analysis but have only recently become feasible for non-motorized travel modes. To date, the bulk of our information on bicycling activity has come from national and regional household travel surveys or observed counts of cyclists -- either short-duration manual or longer-term automated counts -- in a limited set of locations. Based on these datasets, models must be developed to assess network-wide conditions. Direct demand models have been estimated to explain observed counts as a function of surrounding land use, demographics, and other proxies of activity and cycling conditions. As an alternative, bicycle volumes can be predicted as part of a much more complex regional travel demand model, but in practice such models that include bicycling at a useful level of detail remain extremely rare. Recently, new sources of bicycling activity data have emerged. These derive primarily from GPS-based smartphone apps (e.g. Strava), GPS-enabled devices which provide location (e.g. Streetlight) and GPS-enabled public bicycle sharing systems. These emerging data sources have potential advantages as a complement to traditional count data, and have even been proposed as replacements for such data, since they are collected continuously and for larger portions of local bicycle networks. However, the representativeness of these new data sources has been questioned, and their suitability for producing bicycle volume estimates has yet to be rigorously explored.
The research proposed here would develop a method for evaluating and integrating emerging sources of bicycle activity data with conventional demand data and methods, and then apply the results to several locations to predict network-wide bicycle volumes. Anticipated outcomes include:
● literature review, catalog, and evaluation of available third-party data sources and existing applications
● demonstration of bicycle volume models that incorporate emerging data sources
● comparison of the relative accuracy and value added by different data sources and modeling techniques
● openly available scripts and documentation to help others evaluate, process, and apply emerging data sources for network-wide bicycle volume estimation