Modelisation du choix de la gare d'embarquement pour les usagers du train de banlieue accedant en automobile

The number of commuters who access public transit by private automobile (park-and-ride users) has increased over the last twenty years in the Greater Montreal Area. The commuters who access the train by car represent an important number of these travellers. This document presents findings on the behaviour of park-and-ride travellers based on data obtained from 2001 to 2005 commuter rail on-board surveys for the Montreal-Rigaud (5,533 respondents in 2005) and Montreal-Deux-Montagnes (9,965 respondents) commuter rail lines, and a generalized impedance model for the choice of train station. This study deals solely with observed park-and-ride trips; the choice of travel mode is not studied. The goal of this work is to validate and if possible improve on pre-existing models of boarding station choice. As such, the idea is to reproduce observed behaviour rather than to make predictions.;Four parameters are finally considered in the model: access time, parking occupancy level, parking capacity and transit fare. The model is validated with the help of samples of on-board survey responses and the model is calibrated from the observations. The estimation was done using discrete choice modelling, more specifically the clogit (conditional logit) function from STATA. Results show that access time to the train station by car is the most important parameter in the choice of boarding station. However, only 55% of the trips are perfectly reproduced using this calibrated function, a success rate similar to that of a model which assigns park-and-ride travellers to the station nearest their home. The model overestimates the importance of parking capacity since the majority of park-and-ride users choose the largest parking lots. Consequently, there are still improvements to be made at this level. The first improvement would possibly be to separate the two train lines because they are different. Moreover, the hypothesis on the access time and on the transfer time may be called into questions. Finally the four variables may not be the best and other variables could be added in order to obtain more suitable utility parameters.;The number of park-and-ride users increased by more than 20% between 2001 and 2005 for both rail lines. In 2005, the majority of commuters accessed the train by driving their car. The majority of park-and-ride users are women and workers. These travellers are regular commuters who routinely take the train. The large majority of park-and-ride users get off the train downtown. After getting off the train, users mostly walk to the final destination since the terminal stations are in central downtown, but transit is used more and more. Users are looking to minimise their disutility due to the choice of train station. The datasets are so also used to estimate a disutility function for the choice of a boarding station. This choice is relatively complex because it can depend on individual attributes (age, gender, home location, car ownership) and choice attributes (access time by car, occupancy level of the parking lot, parking capacity, parking fare, walk time, waiting time, fare, travel time in the train, train occupancy level). First, only choice parameters are included in the impedance function. In the case of Montreal, we are able to simplify the impedance function. The waiting time is indeed removed because park-and-ride users base their arrival time at the station on their train schedule. In addition, parking is free at all stations, so parking fees are omitted. The assumption of free-flow speed travel is retained for the estimations of access times. Results show that most of park-and-ride commuters (77% in 2005) will board the train at one of the two nearest station. The train occupancy level and the walk time do not seem to have an important influence on station choice. The transit fare is considered since about 20% of park-and-ride users choose a train station located in a lower fare area. The average parking lot occupancy is lower for park-and-ride users who do not choose one of the two train stations that are the nearest to their point of origin. So the parking lot occupancy level seems to have an influence on station choice. The parking capacity does not seem to have the same effect, but stations with high capacity attract people who live further.