Developing an Approach to Modelling and Appraising Reliability That is Consistent Across Modes

Developing an Approach to Modelling and Appraising Reliability That is Consistent Across Modes


R Batley, D Watling, J N Ibáñez, ITS, University of Leeds, UK


We develop, from first principles, a unified approach to modelling reliability across modes. Deriving existing models (e.g. ?mean-variance?) as special cases, we illuminate potential inconsistencies in current practices.


The Eddington study (DfT, 2006) of transport?s role in the economy will have a lasting impact on the way in which transport investment is appraised in the UK. Among other issues, the potentially beneficial impact on reliability was identified as a key component that is typically missing from current appraisal methods, echoing similar concerns expressed in other countries. The study also highlighted the need for consistent approaches across modes. Presently, the reliability methods available and recommended (in the UK at least) do not appear to be consistent across modes. This apparent inconsistency causes particular difficulties in many situations, for example when:
? comparing benefits of two policies that act on different single modes, e.g. rail vs. road;
? evaluating policies that have an impact on modal choice;
? modelling problems in which multiple modes are used as part of a journey, e.g. car as an access mode to train or air;
? transferring insights about behavioural response to unreliability across mode boundaries, e.g. why should an individual?s attitude to time risk fundamentally differ in car vs. train?
The contribution of the paper is to first critically review the methods that have been developed within modes for analysing such problems, and then to present a generic modelling approach that is able to encompass such mode-specific approaches as special cases.

The generic modelling approach to be presented has various elements to its structure, which might be summarised as follows:
? a representation of inter-modal networks for complete journeys, in a way that allows differential levels of spatial aggregation to be included as desired (whether between modes or between, say, the urban/inter-urban scales);
? a representation of ?decision-making units? that may, at one extreme, be individuals or individual households, and at the other extreme be aggregate share models at the origin/destination/market-segment level (with data appropriate to the level of resolution);
? a flexible representation of the decision dimensions, including the choice between different modes and mode-combinations, choice of service/route and choice of departure time;
? a representation of the distinct supply-side characteristics of the modes (e.g. timetabled departures versus continuous departure), including the variability in travel times across trip legs/segments/links;
? a representation of the perceptual connection between actual travel time variability and risk, and the subjective perceptions of summary measures thereof; and
? a representation of decisions in accordance with expected utility maximisation

Having specified the generic modelling framework, we then go on to illustrate the approach with examples based on a common problem. The common problem concerns travel on the inter-urban scale, specifically choice of mode and service/route over a joint road and rail network, where road (as well as an alternative) serves as a potential access mode to rail. Based on an underlying stochastic model, data have been simulated at varying levels of aggregation, characteristic to the kinds of survey carried out for each mode (e.g. SP, traffic counts). Where available, existing data have been used to calibrate the simulated problem to realistic conditions. The benefits of two schemes are then evaluated: one a quality improvement on rail (increase in frequency and reduction in in-vehicle time) and one a quality improvement on road (increase in road capacity), both of which might be expected to yield reliability benefits as well as benefits in the mean. The evaluation was carried out under two consistent modelling paradigms within our generic framework: (1) using the conventional approach in rail (based on scheduling/late arrival), but extended to apply to road also (both as an alternative and as an access mode to rail); (2) using an instance of the recommended WebTAG reliability ratio approach typically envisaged for road, but applied to rail also. The parameters of the simulated data were then varied, and the sensitivity of the benefits under the two modelling approaches derived, for each of the two policy measures. In the paper, the (complex) results of these comparisons and sensitivity tests are reported, and recommendations are given for both practice and further research.


Association for European Transport