Planning and policy initiatives at Metropolitan Planning Organizations (MPOs) are dependent on travel demand models, such as activity-based models (ABMs). Although almost all ABM developments in the U.S rely on professional consulting services, many MPOs are concerned about the amount of modeling efforts. The objective of this paper is to provide some insight and alleviate some of these concerns based on the experience gained from the San Diego Association of Governments (SANDAG) ABM development.

First, the paper discusses why ABM is important and what MPOs can benefit from an ABM in regional long range planning. Using the SANDAG ABM as an example, this paper describes the planning of an ABM project with focus on preparation of a RFP, scope of work definition, model platform selection, and scheduling.

Second, the authors demonstrate that data requirement for the ABM development is significant, but it is by no means excessive. To enforce data consistency, rigorous data quality controls are conducted using GIS and database tools. A description of how data quality control is performed is discussed.

Third, the paper describes model estimation, calibration, and validation, and what roles the MPO modeling staff plays in these steps. Another responsibility of MPO modeling staff is to review the models to make sure they satisfy future policy analysis needs and represent true local travel conditions.

SANDAG envisions an integrated modeling platform that is composed of an ABM, a commercial vehicle model (CVM), a land use model, and a greenhouse gas emission model. In this section, the paper describes the data exchanges among these models. Special market models such as visitor, special event, and cross-border models are also discussed.

Next, the authors discuss various project management issues, including coordination with the consultants and the planning staff, local stake holders, expert panel oversight, hardware requirements, open source software development, and documentation.

Finally, the authors discuss lessons learned and future improvement recommendations, including a large GPS-based household travel behavior survey, reliable highway and transit counts, an explicit child escorting model, an auto ownership model sensitive to gas price, and a dynamic traffic assignment handles dynamic tolling more adequately.