Structural Characteristics and Generative Mechanisms of the Urban Bus Bunching Co-Occurrence Network: An Empirical Analysis Based on Exponential Random Graph Models

The service reliability of urban public transit is closely linked to sustainable socioeconomic development and low-carbon urban mobility. Despite extensive studies on transit reliability, existing research predominantly evaluates bus bunching as an isolated, single-route operational anomaly, leaving an important gap in understanding the network-level structural coupling and its potential environmental implications. To bridge this gap, this study constructs a directed bus bunching risk co-occurrence network based on one month of empirical operational observations from Wuzhong City, encompassing 419 physical stops and 33,387 identified bunching events. Methodologically, this study applies the application of the Exponential Random Graph Model (ERGM) to transit reliability analysis, transitioning from traditional independent-variable assumptions to a common probabilistic framework. This approach distinguishes the generative mechanisms driving network tie formation across endogenous network structures, exogenous nodal attributes, and relational similarities. The empirical results indicate the following: (1) The overall network exhibits a low baseline density but strong bidirectional reciprocity (mutual coefficient = 2.995, p < 0.01), demonstrating strong endogenous structural dependencies. (2) Core stops with higher operational loads and multi-line transfer hubs exhibit a heightened propensity for risk co-occurrence and marked homophily in congestion risk. (3) The impact of the built environment demonstrates substantial temporal heterogeneity, with residential demand explicitly driving morning peak risks, whereas the evening peak is characterized by intensified bidirectional congestion resonance. Furthermore, route-segmented analyses identify four distinct spatiotemporal evolution paradigms of delay cascading. By clarifying the generative logic of network dependencies, this study supports an important transition from reactive dispatching to proactive, topology-driven transit governance. Mitigating the stop-and-go platooning patterns induced by these risk paradigms can reduce transport-related greenhouse gas emissions, providing a robust empirical basis for cultivating a resilient, low-carbon, and equitable sustainable urban ecosystem.