Planning and Policy Impacts of Alternative Public Transport Modeling Methods
H Slavin, J Brandon, J Lam, A Rabinowicz, S Sundaram, J Zhang, Caliper Corporation, US
This paper examines the consequences of applying alternative analysis methods in appraisal of public transport improvements in large congested metropolitan regions. In this research, we focus on public transport pathfinding and assignment.
This paper examines the consequences of applying alternative analysis methods in appraisal of public transport improvements in large congested metropolitan regions. In this research, we focus on public transport pathfinding and assignment although some implications for mode choice analysis are also considered.
There is a rather wide variation in the methods that are utilized for public transport pathfinding and assignment around the world and there are also a variety of interesting emerging methods. Amongst ?classical? methods one finds single path, multi-path, and schedule-based shortest path methods. There are also stochastic methods and equilibrium extensions to various static approaches as well as emergent dynamic models and simulation-based approaches.
It is not commonly understood that particular public transport pathfinding and assignment methods have their own signatures and potential biases that can confound objective analysis of system improvements. The presence and magnitude of these biases, however, has largely remained unstudied.
In this research we examine a variety of modeling methods utilizing networks and origin-destination matrices for the New York City Transit system. Modeling this system is particularly challenging due to the variety of transit modes and the numerous alternative routes available to system users. The system is also congested at peak travel periods leading to further complexity in forecasting route utilization and boarding points.
The research relies on a much better data set than has been previously available for the New York City Transit system. These data have been derived from extensive processing of automated fare collection system card swipes. Automated fare collection (AFC) system data offers the promise of providing more information about demand and traveler behavior than has heretofore been available. AFC data are generated every day and, with some effort, can be processed for use in planning and modeling efforts.
Using implementations of alternative methods, we compute informative statistics about system utilization and error levels in matching system counts. We then examine forecasts of particular system improvement projects and study how various benefit measures vary with alternative methods. From this we identify planning and policy impacts that are influenced by different methodological choices.
Association for European Transport