This article describes the methods used to improve the performance of the Greater London Transportation Studies (GLTS) model for forecasting traffic distribution and modal choice. The objectives were to: (a) replace the simultaneous distribution and modal split model by an appropriate hierarchical structure; (b) replace the fixed mode and route cordon crossing trips by an external zoning system and model; (c) replace the all- or- nothing assignment procedure for choice between public transport modes by a probabilistic model; (d) renew the car availability and trip purpose segmentations; (e) replace the empirical deterrence functions of version 1 by analytical functions; (f) incorporate more policy related variables such as parking cost and supply, and company car ownership; and (g) extend the model to cover the full 24 hour period with factors defined to identify travel in particular time periods. The theoretical framework which was set up to investigate the modal choice hierarchy is described, as are the results of investigations that lead to the adoption of a three level hierarchy of choice models: a distribution, a main mode choice, and a public transport choice. The extensive investigation of the model database which led to the identification of the most appropriate segmentations and trip purposes for modelling is also described. The results of investigations into model coefficient stability with zone size are also presented.
Abstract