Estimating Alternative Departure Time Choice Models Using Nested, Mixed, Latent Class, Ordered and OGV Logit, Applying the Most Appropriate Model Within the London Borough of Islington’s Multi Mode Model Which Used Clocktime Assignment, and Investigating



Estimating Alternative Departure Time Choice Models Using Nested, Mixed, Latent Class, Ordered and OGV Logit, Applying the Most Appropriate Model Within the London Borough of Islington’s Multi Mode Model Which Used Clocktime Assignment, and Investigating

Authors

Peter Davidson, Peter Davidson Consultancy, Adam Thomas, Peter Davidson Consultancy, Collins Teye-ali, Peter Davidson Consultancy

Description

This project was an investigation into adding a time interval choice component to an existing multi-modal model of Islington. It sought to identify a suitable model form, investigate the clock time assignment at reproducing travel time at different times of the day and find how well these performed when asked to test various types and levels of policy intervention.

Abstract

This project was an investigation into adding a time interval choice component to an existing multi-modal model of Islington. It sought to identify a suitable model form, investigate the clock time assignment at reproducing travel time at different times of the day and find how well these performed when asked to test various types and levels of policy intervention.

Various model departure time choice models including multinomial, nested, mixed, latent class, ordered and OGEV logit were estimated on the London Area Traffic Survey household interview 2001 data. Preferred arrival times were added to the observed trip records by imputation using a slightly modified version of PATSI, a program written as part of the development of the Department for Transport’s HADES software in the late 1990s.

One departure time choice model was selected and incorporated into the full supply-demand model structure. In modifying the existing multi-modal model to include departure time choice, the assignment component was changed from the traditional peak hour approach to a continuum of 24 half-hour clocktime assignments, modelling traffic throughout the 12 hour day and linking network conditions and trips in progress at the end of one interval to the start of the next. This was shown to be a necessary modification which worked well at providing travel costs for the demand models.

Four example scenarios were used to assess the model’s response to different types and levels of intervention. These showed that the model’s behaviour performed according to a priori expectations. Simple analysis of the choice model outputs, assigned traffic volumes and consumer surplus benefits showed that the inclusion of time interval choice was justified in terms of improving model accuracy and the range of uses of the model. This study verified that the departure time choice could be nested in an existing multi mode model using the inclusive value. It also showed that clock time assignment worked well at measuring the different congestion conditions throughout the day.

This is an important piece of work which contributes to our knowledge of departure time choice modelling in practice - there are not many examples of successfully including departure time choice into an existing working multi-mode model. One of the main obstacles has been the limited scope in assignment models for getting different skims for each time interval throughout the day. Using clock time assignment, we were able to get different skims to a granularity of 10 minutes ie we could get different skims for each of the 72 ten minute time intervals through the 12 hour day.

Publisher

Association for European Transport