Dwell time represents a significant portion of bus operation time and variance. While dwell time is highly correlated to the number of passengers boarding and alighting, there are also secondary factors, specifically crowding, fare type and bus design that may have a significant impact on dwell time. It is these secondary factors that influence the effectiveness of service improvements. For instance, smart media fare cards are estimated to have a 1.5 second faster transaction time than magnetic strip tickets, but only in non-crowded situations. When the number of on-board passengers exceeds the seating capacity, there is no statistically significant difference between the fare media types. Automatic collection systems provide a plethora of data but require preprocessing to combine records from different collection systems, to control for measurement error, and to determine the significant factors of dwell time. Using data from the automatic passenger counting (APC), automatic fare counting (AFC), and automatic vehicle location (AVL) systems of the CTA bus network, I develop and implement preprocessing techniques, estimate a dwell time model, and analyze the impact of the secondary factors.
Abstract