Modelling local rail demand in South Wales

Abstract

Direct demand models have been developed based on ticket sales data for 85 local rail stations in South Wales. Initially log-linear regression models were calibrated, incorporating a variety of independent variables. Geographical Information Systems were used to implement flexible station catchment definition methods, notably flow-specific catchments where population units were allocated to stations by minimising the total travel time to individual destinations. To validate these methods a survey of ultimate trip end locations was carried out on the Rhymney line in South Wales. To give consistency with predictions from trip end models, methods were developed to constrain predicted trip numbers for each flow based on the total trips observed or predicted from origin stations. Simple scaling had only limited success, so probabilistic trip distribution models were calibrated by using both the linear and non-linear regression. These gave superior results and explicitly incorporated the effects of intervening opportunities in the model form. © 2010 Taylor & Francis.