Decomposition and Sensitivity Analysis of Bus Travel Times

Transit service reliability is important for transit planning and operations as well as passenger experience. Large travel time variations increase operating costs and negatively affect passenger satisfaction. Existing literature focuses on specific aspects of transit travel times but less on how these aspects interact with each other. This paper proposes to combine previous research efforts by further decomposing observed trip travel times into four elements using 3 months of archived vehicle location and fare transaction data. Departure times and inter-stop travel times are obtained from vehicle locations. Dwell times at stops are estimated from fare transaction data using a dwell time model. Red-light waiting times are calculated using the vehicle locations and estimated signal timing plans. Then, using these as inputs, we identify important trip elements affecting the overall travel time variation, as well as how much variation can be attributed to each trip element using variance-based and one-at-a-time sensitivity analyses. The overall travel times and red-light waiting times are more affected by interaction effects between trip elements, whereas the overall inter-stop times and dwell times are mainly affected by large individual variations. The results suggest that planners must consider potential chain reactions where small variations in one trip element can lead to significant changes in the overall trip times as a result of interaction effects with varying cycle lengths in fixed signal timing plans. These findings will help planners better integrate available data sets, carry out comprehensive analyses, and pinpoint the determinants affecting travel time variation on each route.