Temporal Transferability of Mode-destination Models: Summary of Literature, Initial Findings
WINNER OF The Neil Mansfield Award
J Fox, ITS, University of Leeds, UK
This paper summarises progress to date on the author's PhD research into the temporal transferability of mode-destination models. The findings from the literature review are presented, together with findings from initial analysis of Toronto data.
This paper presents progress to date on the author's PhD research into the temporal transferability of mode-destination models.
The motivation for the research effort is that the issue of model transferability is important in the context of transport planning, where forecast horizons of up to 30 years are used. Implicit in such applications is the assumption that travellers tastes, as represented by the model parameters, are constant over time. The research effort seeks to investigate the existing evidence base underlying that assumption, undertake additional empirical analysis to add to the evidence base, and set out recommendations for practitioners to improve practice.
The paper will set out exactly what is meant by model transferability, drawing a distinction between temporal and spatial transferability. Measures that can be used to assess model transferability will be detailed, with a discussion of their relative strengths and weaknesses.
The next section of the paper will summarise the findings from the literature review (recently published in TRR 2175). The review highlights how most of the evidence dates from early applications of discrete choice transport models in the late 1970s and early 1980s. It also highlights that the evidence is concentrated on mode choice studies applied over short time periods (typically a few years), and that little empirical evidence exists as to the transferability of mode-destination choice models over longer term forecasting horizons.
Next, the initial findings from the new empirical analysis will be presented. Toronto Transportation Tomorrow Survey (TTS) data collected in 1986, 1996, 2001 and 2006 will be used to undertake tests of model transferability over transfer periods of up to 20 years. The initial analysis will develop base models from the 1986 TTS data, and assess their transferability over 10, 15 and 20 year horizons using the 1996, 2001 and 2006 TTS datasets, together with supporting level-of-service and land-use data for each year. The analysis will be undertaken using home-work tours.
The paper will conclude with findings from the literature review and initial empirical analysis, and then set out the plans for the remainder of the research effort.
Association for European Transport