An investigation is reported on bus bunching and its spatiotemporal patterns that considers six bus routes in Kolkata City, West Bengal, India. With historical GPS data as inputs, bunching occurrences are identified using the spatiotemporal-density-based spatial clustering of applications with noise (ST-DBSCAN) algorithm. To determine the impact of bunching on various bus stops, k-means++ clustering was used on top of ST-DBSCAN, and subsequently, spatiotemporal bunching patterns were investigated. Some bus stops are consecutively impacted by bunching; however, others are impacted in an isolated manner. To capture the spatiotemporal propagation of bunching, a new indicator termed the propagation ratio is proposed, which could classify the bunching as propagating or isolated. Bunching occurrences are propagating for some bus routes and isolated for others, which justifies the need to prioritize routes displaying propagating nature. Of interest, bunching is identified on high and low-frequency bus routes, which indicates the need to relook at the operational planning of all the bus routes regardless of their frequency. At times, excessive bunching is detected during lean periods of the day, which suggests the necessity of modifying the dispatching headways that are based on passenger demand during these lean periods. The outcomes from this work further indicate that all the bus stops along a bus route are not equally impacted by bunching. The spatiotemporal cluster plots could successfully capture the initial bunching occurrence points and their variation with time, which could provide an opportunity for the operators to apply time-varying remedial measures. The insights developed from this work could be used advantageously by operators and policymakers for mitigating bus bunching and improving the urban bus service.
