Defining probability - based rail station catchments for demand modelling

The aggregate models commonly used in the UK to estimate demand for new local rail stations require the station catchment to be defined first, so that inputs into the model, such as the population from which demand will be generated, can be specified. The methods typically used to define the catchment implicitly assume that station choice is a deterministic process, and that stations exist in isolation from each other. However, studies show that predefined catchments account for only 50-60 percent of observed trips, choice of station is not homogeneous within zones, catchments overlap, and catchments vary by access mode and station type. This paper describes early work to implement an alternative probability-based approach, through the development of a station choice prediction model. To derive realistic station access journey explanatory variables, a routable multi-modal network, incorporating data from OpenStreetMap, the Traveline National Data Set and National Rail timetable, was built using OpenTripPlanner and queried using an API wrapper developed in R. Results from a series of multinomial logit models are presented and a method for generating probabilistic catchments using estimated parameter values is described. An example probabilistic catchment is found to provide a realistic representation of the observed catchment, and to perform better than deterministic catchments.