Modelling probability distributions of public transport travel time components

Abstract

Understanding the variability of public transport travel times is essential for various reasons, e.g. for gaining knowledge of the deteriorations and ameliorations in daily traffic, for providing adequate (real time) information to customers and for optimizing transit schedules. This paper deals with this issue by modeling the day-to-day variability of travel times of urban buses based on planned and actual arisen arrival and departure time data of selected bus routes in Zurich by the aid of statistical distributions. For a detailed description of the variability, the travel times are decomposed in running and dwell times. The most appropriate distribution models for the running, dwell and travel times at different times of the day are identified, and their performance is evaluated. This is done under consideration of the issue of how the temporal aggregation (i.e. length of departure time window) as well as the spatial aggregation (i.e. number of considered bus stops or sections) influence the nature and shape of these distributions.