Spatiotemporal analysis of traffic oscillation propagation at consecutive merging bottlenecks using trajectory data

Abstract

Spatiotemporal traffic oscillations at consecutive merging bottlenecks play a critical role in the formation and amplification of congestion. A better understanding of their propagation is essential for developing effective traffic control strategies. While previous studies have primarily examined oscillations from macroscopic perspectives or in isolated bottleneck settings, the interactions between oscillations across closely spaced merging bottlenecks remain underexplored. This study addresses this gap by analyzing empirical trajectory data to investigate oscillation dynamics at a mesoscopic, lane-specific level. To accurately identify oscillation patterns, a wavelet transform method was applied to the trajectory data. The data were then aggregated using a spatiotemporal window structure. The Spatial autoregressive with autoregressive disturbances (SARAR) model, incorporating a Gaussian kernel-based spatiotemporal weights matrix, was employed to quantify the influence of both endogenous and exogenous factors. Results found that traffic oscillations exhibit a significant spatial spillover effect, whereby an existing oscillation can trigger a new oscillation in adjacent regions. However, this spatial effect is not deterministic; theoretical lag distance analysis shows that local disruptions can be absorbed if vehicles maintain stable operation for a sufficient duration (approximately 100 s in our data). Furthermore, the upstream bottleneck acts as a powerful amplifier, inducing a unique oscillation pattern. Merging events within these oscillations exhibit a distinct characteristic: both pre- and post-merge time headways remain consistently low (around 3 s). This persistent low-headway merging serves as a mechanism for cumulative disruption in acceleration and speed. This study provides a framework to characterize spatiotemporal propagation of oscillations and highlights the importance of local interactions in congested traffic flow.