The phase 3 Arizona statewide model (AZTDM3) development phase incorporates a mode choice capability to estimate transit and rail passenger trips from person trip tables. The mode choice model will account for longer-distance, high capacity service as well as shorter-distance, intra-urban transit services. The planning requirements for the model are to produce reliable mode share forecasts of large scale transit and rail feasibility studies. In particular, this model is designed to be used for the Phoenix- Tucson high speed rail feasibility study. This paper focuses on two particular aspects of the mode choice study: transit abstraction methodology and long distance passenger travel models.

Coding public transit networks is a difficult and time-consuming task for statewide travel demand models. Such models often do not forecast inter-city transit usage, because there is insufficient zonal and network detail to produce reliable estimates. While micro coding of public transit networks is not necessarily applicable in a statewide modeling context, it is necessary to consider the same to properly account for accessibility to and from the local transit system to the regional/intercity system. For a statewide model like AZTDM3, a logical way to overcome this difficult task is to estimate aggregate models to represent the local transit system. These models, known as transit abstraction process which were implemented in California, are being calibrated and validated based on the actual transit characteristics recorded in National Transit Database (NTD) and Section 5311 statistics. This is the first application of transit abstraction outside California which serves as a case study for other statewide models looking to implement a similar strategy.

Long distance travel models are an essential component of statewide models distinguishing them from other urban and regional models. 1995 American Travel Survey (ATS) data has been used to estimate a county to county long distance person trip table for the entire country. From this database, a TAZ to TAZ person trip table matrix has been obtained using data from the 1995 ATS, and the 2001 NHTS. Nested logit models with ascertained coefficients from other statewide models and high speed rail studies (such as Quebec-Windsor, Orlando-Tampa) have been used to further obtain trip tables by purpose (business , non-business) and mode (drive alone, shared ride2, shared ride 3+, walk to transit and drive to transit).