Assessing the Impact of Transit Right-of-Way on Service Reliability via Segment-Level Data Integration and Ensemble Learning

Abstract

Transit service reliability is a critical determinant of passenger satisfaction and system efficiency. While dedicated transit rights-of-way (ROW)—such as bus lanes and busways—are known to improve travel-time reliability, most existing studies are corridor-specific and do not generalize to system-wide planning. The analysis is often at the route level, ignoring potential reliability variation along routes. Moreover, the effects on travel-time variability of alternative ROW treatments, such as bus-on-shoulder, high-occupancy vehicle (HOV) lanes, and high-occupancy toll (HOT) lanes, remain underexplored. This study addresses these gaps by deriving segment-level reliability metrics using high-resolution automatic vehicle location data and automatic passenger count data for the entire transit network in the Minneapolis-St. Paul Twin Cities metropolitan area. A gradient boosting regression tree model is employed to evaluate how various ROW types, service characteristics, traffic conditions, and land use features affect travel-time variability. Results show that substantial reliability improvements occur only when over half of a route segment is dedicated to bus running, highlighting the limited impact of partial ROW implementations. The study also finds that bus-on-shoulder and HOV/HOT lanes offer limited reliability benefits under non-peak conditions. In addition to ROW, other factors, such as signal density and operating environments along route segments, also significantly affect their travel-time variability. The trained model can support scenario analysis and guide ROW planning by estimating the impacts of specific implementations and helping prioritize investments based on projected user benefits and reliability gains.