Many cities, especially in developing countries, are experiencing a serious imbalance between an explosively growing demand for space and mobility and limited resources in terms of land and transportation This imbalance is particularly exacerbated during crowding events. Resulting crowd disasters have witnessed nearly 3,000 fatalities in the last five years, all over the world.
Crowding-zone specific, microscopic models of pedestrian behavior were recently studied to find the rules for designing better space geometries and achieving better crowd organization. However, such crowding-zone specific models lack a human mobility component covering the phase preceding the actual crowding event. This is very important, as it is actually difficult to evacuate people both safely and rapidly when high-density crowds have already formed.
Here, we answer the important question of how crowds come together before they enter the crowding zone in the first place. We use human mobility data collected during the last three months of 2014 via nearly 200 million individual trip records based on more than 6 million subway smartcards and 13,731 taxies in Shenzhen, a major city in China. The long observation period and the accurate time and coordinate records guarantee that the mobility fluxes during crowding events can be traced at an unprecedented accurate and comprehensive level.
Counterintuitively, during the 14 confirmed crowding events in our data, mobility fluxes, trip travel time, trip distance and duration of stay at the event location show no obvious difference from regular behavior. Put differently, dangerous crowding situations are hard to detect by monitoring urban traffic.
We introduce the concept of anomalous mobility networks, where, nodes are traffic zones, and a link is generated between two nodes if the mobility flux between nodes exceeds the upper threshold defined via an algorithmic measure (the Jensen-Shannon divergence), comparing the present state with historical records. Surprisingly, we find that a simple topological measure of anomalous mobility networks, the in-degree of a crowding zone or node, k_in, that is not the volume of traffic, but the diversity of traffic sources, can predict large crowd gatherings at early stages.
Interestingly, in anomalous mobility networks, the majority of links connecting to the event zones (or nodes) possess low flow volume. Nevertheless, via the k_in-measure, anomalous mobility networks can accumulate evidence for potential crowd gatherings using low-intensity connections. This finding will make lasting impact in crowd disaster prevention and urban transportation, due to low-volume flows previously being hard to capture and therefore usually ignored.
In summary, the present study uncovers and helps to easily identify essential mobility patterns feeding into large crowd gatherings. Our study contributes valuable, likely life-saving insight to anticipate and avoid potentially dangerous crowding situations. Most importantly, the transportation data used is easy to collect, widely available, and preserving the anonymity of citizens, allowing for broad global acceptance of the proposed approach.
Huang, Z., Wang, P., Zhang, F., Gao, J., Schich, M., 2018. A mobility network approach to identify and anticipate large crowd gatherings. Transp. Res. Part B Methodol. 114, 147–170. https://doi.org/10.1016/j.trb.2018.05.016
Guo, B., Yang, H., Zhou, H., Huang, Z., Zhang, F., Xiao, L. and Wang, P., 2022. Understanding individual and collective human mobility patterns in twelve crowding events occurred in Shenzhen. Sustainable Cities and Society, 81, p.103856.
Zhou, H., Zheng, Z., Cen, X., Huang, Z. and Wang, P., 2021. A data-driven urban metro management approach for crowd density control. Journal of advanced transportation, 2021.