Modelling Passenger Crowding in London: Testing of Alternative Methodologies and Assumptions



Modelling Passenger Crowding in London: Testing of Alternative Methodologies and Assumptions

Authors

M Swiderski, SKM Colin Buchanan, UK

Description

Transport for London recently commissioned the testing of Capacity Constrained Transit Assignment (CAPTRAS) within its strategic public transport model with the purpose of evaluating this method against its standard crowded assignment.

Abstract

Transport for London's (TfL) strategic public transport model, Railplan, currently takes passenger crowding into consideration by adding an in-vehicle crowding "penalty" to the in-vehicle service time, the Congested Transit Assignment (CONGTRAS) method. The relationship of the size of these penalties to the level of capacity available on the services was determined through survey work undertaken in the 1970s in which the preference of passengers to either board a service or wait at the platform was recorded.

TfL recently commissioned SKM Colin Buchanan to test an alternative methodology within Railplan, Capacity Constrained Transit Assignment (CAPTRAS), which adjusts modelled passenger wait times through application of a service headway function, with the size of the adjustment being determined by the capacity available at the public transport stop, after taking into account passenger boardings and alightings at the stop. TfL wanted to understand the possible benefits as well as challenges/risks associated with any implementation of CAPTRAS within Railplan and required the consultant to provide recommendations for when and if each method should be applied within the various uses of Railplan.

Initially, INRO (the developers of the EMME software within which Railplan operates) updated the CAPTRAS program to purpose-fit Railplan allowing a pure comparison to be undertaken between the CONGTRAS Railplan and CAPTRAS Railplan assignment methods.

Testing of CAPTRAS subsequently provided some evidence of marginal improvements in model validation. This was noticeable firstly after adjusting the beta value within the headway function, which determines the sensitivity of demand flows to changes in capacity available for passengers. Secondly, it was noticeable after reducing assumed public transport service capacities to 5 passengers per square metre of standing space, which is more in line with passengers? perception of capacity limits, from the standard 7 passengers per square metre of standing space, which is based on physical crush capacity limits.

However, more detailed analysis of individual sections of railway which are known to be crowded in the "passenger perception sense", were found to have an increased over-estimation of demand relative to observed passenger flows. This was somewhat unexpected, given that use of the effective headway function entails greater consideration of the limited capacity of services both in terms of the application of increased wait times during model assignment and in terms of model convergence, which requires demand on all service segments to fall within capacity limits or for excess demand to be minimised according to user specification. This implies that CAPTRAS shows that these sections of railway are not as crowded as passengers perceive them to be. The reason for this discrepancy is that the effects of uneven loadings along the length of the train, and of peaked demand profiles within the modelled period that are not fully matched by similar peaks in service capacity, are not fully taken into account within CAPTRAS. Therefore, more work is required to test alternative beta values and service capacity assumptions used in CAPTRAS, in order to fully utilise its potential to improve the model's forecasting of demand and crowding levels.

Another issue arising from the work was of a more theoretical nature. The in-vehicle crowding penalty represents passengers' perception of crowding modelled as a time penalty. Whilst the headway function models passengers' actual wait time, the size of the adjusted wait times indicates that they must also partially act as a penalty rather than simply being a realistic representation of the actual time passengers would be willing to wait for a service. Therefore when using the in-vehicle crowding penalty and headway function in combination, there should be some consideration as to whether the crowding penalties are being double-counted. In a similar vein, it could be argued that the realistic increases in modelled wait time might already be included within the in-vehicle time penalty. It is important to resolve this issue in order to not exaggerate the effects of crowding on demand levels and to not double-count passenger time benefits as part of economic appraisal of new transport schemes.

It is also interesting to evaluate the crowding assignment methodologies above with that adopted by the Department for Transport (DfT) in their modelling. A similar approach is used to the CONGTRAS approach adopted by TfL, whereby an in-vehicle time penalty representing passenger discomfort associated with crowding is applied. The main difference lies in the DfT?s use of elasticities to determine the sensitivity of demand to changes in crowding.

Publisher

Association for European Transport