The Continued Innovation of Aggregate Transport Demand Models

The Continued Innovation of Aggregate Transport Demand Models


N Benbow, P Kidd, A Woolley, MVA Consultancy, UK; A Skinner, ACS Transport Consultants, UK; I Palmer, Greater Manchester Passenger Transport Executive, UK


Tiered aggregate demand and detailed network models will be increasingly valuable tools for policy makers in conurbations. Advances have been made in the consistency between model tiers in work to support multi-billion pound investment proposals.


Aggregate transport demand modelling continues to evolve with state-of-the-art tiered modelling systems now being developed. Applications of this type have been used to provide supporting evidence for the high-profile, multi-billion pound Transport Innovation Fund (TIF) bids of the large conurbations of Greater Manchester and the East Midlands. Such models are likely to play a continued role in the modelling of large conurbations. These models have provided robust demand forecasts and appraisals of a broad range of policy and investment options including road user charging, major highway and public transport infrastructure investment and Smarter Choices. Competing requirements to meet political timetables within budgets, whilst providing robust analyses, have imposed constraints on the modelling approach.

Demand segmentation and network representation have become more complex over the last decade taking advantage of the rapid advancement in computing; both processing speed and memory have increased dramatically. Aggregate transport modellers have taken full advantage of these developments seeking to produce models with ever finer zone systems, segmenting demand by more journey purposes, income, household car availability and time periods and increasing the complexity of network representations to allow more accurate generalised costs to be derived for input to the demand model. It is still not possible to combine within a single aggregate five stage modelling application the levels of demand segmentation and model functionality envisaged by the early pioneers of the approach, with comprehensive highway and public transport network representations of a large conurbation. A decade ago the land-use transport interaction model of Greater Manchester had only 50 zones covering the County, a highly abstract network representation making use of area speed/flow curves, and so was only appropriate for indicative strategy testing. The latest model has 250 zones and has a comprehensive representation of transport networks. A 250 zone system is still coarse for an area with a population of 2.5 million. This coarseness and model run times preclude incorporating representation of traffic interaction at junctions, and the resulting delays, in as sophisticated a way as some mainstream highway assignment packages. However, these challenges can be addressed using a tiered approach.

The tiered model system developed for the two TIF bids has a lower tier containing spatially disaggregate highway (SATURN) and public transport (TRIPS) models that comply with UK Department for Transport (DfT) validation requirements. The upper tier consists of a bespoke aggregate 5-stage transport model known as TRAM which is consistent with DfT guidelines. For example the ratio between numbers of zones in the two tiers for Greater Manchester is approximately 3:1. TRAM takes as input turn flow-delay curves derived from SATURN together with public transport inputs from TRIPS. Demonstrating consistency between the model tiers, through highway journey time and public transport generalised cost comparisons, is crucial to ensuring the validity of forecast and appraisal results. A vital element in this modelling approach involves demand forecasts being passed down from the upper tier TRAM model to the spatially disaggregate SATURN and TRIPS models. Subsequent assignment in these detailed models provides more accurate highway and public transport link flows and travel times to inform a range of appraisals.

Whilst significant advances have been made in the development of tiered models further progress would be facilitated by changes to software architecture. A blueprint is proposed for future development, which would integrate specialised highway and public transport assignment packages, mainstream IT databases, and compiled programmable code for implementing demand modelling. This would be a truly innovative aggregate demand modelling system. Such a system would readily take advantage of future developments in mainstream IT and transport software, being fully interoperable with standard IT data formats.

In conclusion, whilst aggregate models will continue to play a major role in transport policy and scheme appraisal, today?s computers do not yet have the power to integrate detailed supply representations with an exhaustive segmentation of travel demand. For this reason tiered model systems are likely to become more common for policy and scheme appraisal in large areas. If the results of these systems are to offer reliable analyses for policy makers it is imperative that consistency between tiers can be assured. Advances have been made in this regard for work to support bids for multi-billion pound transport investment programmes accompanied by congestion charging. Innovation in software architecture will facilitate further improvements in integrating model tiers, and would allow modellers and their clients to benefit from developments to mainstream transport modelling and database packages.


Association for European Transport