Aggregation of Urban Areas in Inter-urban Traffic Models
GORDON A, Transport Research Laboratory, UK
Recent years have seen an increase in the number of inter-urban regional traffic models that have been developed in the UK. In the early 1990s these were typified by models such as the South-East Regional Traffic Model (SERTM) and the Eastern Region Traff
Recent years have seen an increase in the number of inter-urban regional traffic models that have been developed in the UK. In the early 1990s these were typified by models such as the South-East Regional Traffic Model (SERTM) and the Eastern Region Traffic Model (ERTM). These are assignment-only models that have large study areas 150-200km across. Urban areas are represented in skeletal detail, with just a handful of zones and links and no explicit junction modelling.
This paper describes work comissioned from TRL by HETA division of te Department of the Environment, Transport and the Regions. Models such as those described above provided the motivation for the project. There was concern that the skeletal representation of urban areas was mis-representing the extent of travel time delays caused by traffic in urban areas. Although these models are mainly concerned with inter-urban traffic, urban delays have a significant impact in inter-urban route choice, as well as other travel choices such as mode and departure time.
Part of the project involved an investigation into alternative ways to represent urban road networks with the aim of increasing the accuracy of modelled urban travel times in inter-urban models. While it is technically feasible to carry out detailed junction modelling of every urban area in the model (the NAOMI model, which covers a similar area to SERTM, models a vry large number of urban junctions), there are advantages to developing a more aggregate approach:
* Reduced data collection and maintenance - collecting detailed link and junction data for several urban areas is an onerous task; even where existing urban models are available the data is likely to be of variable age and quality and to be for different modelling packages (e.g. SATURN, TRIPS or CONTRAM).
* Improved convergence - in general, the more detailed the network and the more junction modelling there is, the more difficult it is to achieve acceptable levels of convergence.
* Reduced model run times - current inter-urban traffic models with detailed junction modelling typically take around 48 hours to converge, although some take much longer. This limits the number of scenerios that can be tested and can make it impractical to include the model in an iterative loop with a demand model.
Of course there are also disadvantages of aggregation, although their relative importance will depend on the context in which the model is to be ued:
* More difficult to model small-scale urban network changes - an example might be the pedestrianisation of a town centre street; however, inter-urban models are not intended to appraise such schemes.
* Less accurate modelled link flows in urban areas - again, inter-urban models are not concerned with urban link flows per se, but with urban travel times. However, it is important to ensure that peri and inter-urban flows are modelled correctly.
* Less accurate modelled travel times - this is the main question mark over aggregate networks in the context of inter-urban models. When comparing different aggregation methods the accuracy of modelled travel times is one of the key criteria by which they are judged.
A number of different aggregation methods were applied to a detailed SATURN simulation netowrk. These methods are explained in the next section. Results concerning total travel times, origin-destination (OD) travel times, and link flows are presented to provide a basis for comparison between methods. From these results it has been possible to draw a number of conclusions about the particular methods, and network aggregation is general.
Association for European Transport