Abstract:
Multimodal integration is an important issue in public transport systems due to its influence on both passenger experience and overall network efficiency. In most countries in the global South, achieving integration is particularly problematic because of the informal nature of most public transport. Decentralised service planning and demand responsiveness lead to often uncoordinated, highly variable service patterns, which are not optimised from a passenger perspective. Efforts to promote integration are also hampered by a lack of planning data on routes, service frequencies, and transfer locations. This research asks whether GPS data supplied by passengers as they move through the network can be used to help form a better understanding of the extent and quality of the transfer experience. The data was collected in the City of Tshwane, South Africa, among informal minibus-taxi passengers. Post-processing involved the use of a machine learning algorithm to identify in-vehicle, wait and walk segments, which were used to identify transfers between one vehicle and another. The results showed that many transfers are spatially efficient with short walk and wait times, but that a minority of transferring passengers may experience very long transfers. Transfers encompass a diverse range of behaviours including pacing, shopping and browsing, and typically involve much more walking than waiting. Transfers also occur across a wide range of locations, but tend to be concentrated in certain nodes and along street segments. Strategies to improve transfer facilities as well as general walkability might be targeted at such locations. The study demonstrated that volunteered GPS data is a promising source of information to help planners understand the transfer experience in multimodal networks in data-poor environments.
