Use of logit averaging for integration of urban and intercity passenger rail demand forecasts
Corresponding Author: Lucile Kellis, Steer Davies Gleave
Presented By: Lucile Kellis, Steer Davies Gleave
Abstract
Analyses of demand forecasts for high-speed and intercity passenger rail (HSIPR) service are often carried out using regional / statewide / corridor-level or other models that treat urban areas within the study area at much coarser levels of detail than the MPO or urban models do. Such long distance intercity models typically use simplified representations of local transit accessibility and connectivity near the HSIPR stations, making improvements to them hard to represent in intercity modeling. In addition, the norm for these models is to use the travelers’ choice of HISPR stations in a deterministic manner contrary to real life practice. By the same token, improvements in intercity service or increases in ridership are often not well represented in urban models. It is rare to see urban and intercity models at these two levels to be highly integrated with any kind of analytical rigor. And such an integration is usually not a high priority when conducting customized HSIPR demand forecasting. The cycle was about to be repeated in Texas, where analyses of intercity HSIPR between major metropolitan areas required exchange of information between MPO and corridor-level intercity travel forecasting models. Rather than following the traditional practice of keeping urban and intercity models separate and trying to patch the two together in modeling post processing steps, we built a highly flexible method of exchanging mode choice logsums between the urban and intercity models, making each appropriately sensitive to those calculated using the other. This is achieved through a novel logit average technique that can be easily transferred and applied in other locations with modest changes to the modelling systems used at each level. Moreover, an efficient by-product of this innovative approach is the ability to obtain probabilistic rather than deterministic HSIPR station assignments for the study area travel analysis zones.