Enhancing station level Direct-Demand models with Multi-Scalar accessibility indicators

Direct-demand models (DDM) are increasingly being used for a diversity of transit research and practice purposes. Yet few station-level DDM studies have explored the use of composite indicators of metropolitan accessibility in predicting demand. After all, provision of access to metropolitan destinations is one of the main goals of rapid-transit systems. Furthermore, to this author's knowledge no study has explored potential interactions with local-level accessibility indicators that are typically included in station level transit DDMs. This study explores these possibilities and uses Los Angeles multimodal rapid-transit network as a representative case study of a system that operates in a dispersed agglomeration where multiple sub-centers are linked. Multi-level generalized linear models were implemented where key predictors, including stations' metropolitan- and a local-accessibility indicators are regressed onto average weekday boardings. Furthermore, more general accessibility constructs were developed via EFA and implemented in models; and parameters non-stationarity was assessed via geographically weighted regressions. Results indicate that nodal metropolitan accessibility is a significant predictor of patronage in LA's rapid-transit network, and that its interaction with local-accessibility amplifies boardings and improves DDM models’ explanatory power. More general constructs of accessibility at metropolitan and local-scale were derived via EFA and these resulted in a more parsimonious model with equal predictive power. Land-use and transit planners would benefit from including an accessibility lens in their DDM modeling. Practical applications of these type of models include TOD scenario planning, comparative route alignment studies, system expansion studies, and for didactic purposes given the ability of accessibility measures to capture land-use/transportation interactions.