A regional transportation model, also known as a regional travel demand forecasting model, is a mathematical representation of the supply and demand for travel in an urban area. The travel supply is generally represented by a highway network and a transit network. The highway network represents all major roads in the region and the transit network represents all public transportation service in the region, such as bus, Metrorail, and commuter rail. In addition to transportation networks, the other major input to the travel model is the land activity data for each transportation analysis zone (TAZ). The demand for travel is developed using a series of mathematical models. The most common paradigm for travel demand models in the U.S. is known as the "four-step" model, due to its four main steps:
- Trip generation
- Trip distribution
- Mode choice
- Trip assignment
The four-step model is a trip-based model that is used, in one form or another, by the majority of Metropolitan Planning Organizations (MPOs) that perform regional travel demand modeling. Nonetheless, some MPOs have started to develop activity-based travel demand models, or ABMs, which are discussed on the four-step model page. The first three steps of a trip-based travel model are used to estimate the demand for travel. In the fourth step, trip assignment, the travel demand is equilibrated with the travel supply, as trips are loaded onto one or more transportation networks. The geographic unit of analysis in a four-step model is the transportation analysis zone (TAZ). Usually an urban area is divided into hundreds or thousands of transportation analysis zones.
The TPB travel forecasting model has a modeled area covering 6,800 square miles, including 22 jurisdictions/counties. This area is divided into about 3,700 zones. A typical highway network used by the travel model includes about 45,000 links (road segments) and a typical transit network includes over 600 transit routes, including such modes as Metrorail, Metrobus, other local bus, commuter bus, and commuter rail. Depending on the application, each model run requires about 30 hours of processing time on a high-end workstation. Additional processing time is sometimes necessary for modeling special scenarios.
The cost of maintaining and applying the models constitutes almost half of the region's transportation planning budget, including data support.