AbstractsTransportation

Public transport is known as a sustainable transportation mode. In order to increase demand and cost effectiveness much attention is paid to increase the service quality. One of these service quality aspects is the travel time reliability of a public transport service. However, the monetization of travel time reliability is still a topic of debate since it is unknown how travel time reliability (TTR) should be measured. The implementation of travel time reliability in the route choice behaviour of transport models is still not very common in practice. This research considers five model alternatives (or generalized cost functions), all based on literature, which are estimated from observations of a stated preference survey. These model alternatives mainly differ in the use of the TTR indicator; standard deviation of travel times and the reliability buffer time. The latter is defined as the amount of time a passenger has to take into account to arrive on-time in 80% of the trips. From the model estimation it is observed that it might be worthwhile to estimate a latent class model, since multinomial logit models segmented on observable characteristics or models with randomly distributed parameters, were not able to significantly estimate TTR related parameters with the correct sign. An incorrect sign means that the respondents reacted contrary on what is expected; a more unreliable route alternative is more attractive. A latent class analysis was performed, since there was a suspicion that different perceptions exist regarding TTR. This analysis showed that not every individual took TTR into account in their route choice behaviour or they had difficulties in considering TTR, transfer and frequency at the same time in their trade-off. The resulting different perceptions on TTR, estimated in the latent class model, could not be captured by segmenting on observable characteristics, such as trip purpose. From the results of the model estimations it is concluded that a model with scheduled travel time and reliability buffer time, as a TTR indicator, showed the best results regarding route choice behaviour. The preferred model alternative will be implemented in the regional traffic model of Utrecht, called the VRU-model. The implementation of this model is not trivial and many difficulties are encountered. Therefore, several implementation approaches are described.