Travel Demand Models Help Predict Travel Patterns

Travel Demand Forecasting involves the use of a variety of models to duplicate Trip Generation, Trip Distribution, Mode Choice (e.g. car, transit, bicycle, etc.), and Highway/Transit assignment in a metropolitan region. The models are coded using a program language that is typically executed using Cube or TransCAD software. The primary inputs into the model are highway and transit networks for the representative region, and a demographic file that includes households, population, and employment.

The demographic file is used in the Trip Generation model to estimate the number of trips for each subdivided area, or Transportation Analysis Zone (TAZ), in the region. The Trip Generation model calculates trips per household as a function of household size, income, and oftentimes vehicle availability as trip generation rates fluctuate with these variables.

Trip Distribution is typically executed using a gravity model where zones with large concentrations of employment are most attracted to zones with large concentrations of households nearby. Many travel models include a feedback loop where the congested highway times from the assignment model is input into the Trip Distribution model to account for the impacts of traffic congestion on where people live, work, and shop.

Mode choice models use either a logit or nested-logit structure, where the choices to drive, carpool, use toll lanes, use transit, or bike/ped are based on the utility of using the individual modes. Nested logit models in many major metropolitan areas include an additional “nest” in the transit choice that includes the choices of bus, commuter bus, Bus Rapid Transit, Light Rail, Heavy Rail, and Commuter Rail. The output from the mode choice model is a transit trip table and vehicle trip table.

The vehicle and transit trip tables are then assigned to the respective transit and highway networks, typically using an equilibrium assignment procedure where the model reaches equilibrium when no person can improve his or her travel time between the TAZs in the network.

There are a variety of emerging modeling techniques as well as post processing procedures that have been developed to improve the travel model’s ability to forecast dynamically priced toll lanes, bicycle and pedestrian improvements, and ITS strategies. These include:

  • Activity Based Models which are gradually replacing the 4-step (Trip Generation, Trip Distribution, Mode Choice, Assignment) models used by most Metropolitan Planning and State Agencies
  • Mesoscopic models which simulate traffic at a regional or subarea scale during peak periods (though not at the detail of microsimulation models)
  • National Cooperative Highway Research Program (NCHRP) techniques and associated post processing tools used to develop intersection/interchange forecasts, and
  • Other post processing techniques to evaluate bike/ped improvements, and Travel Demand Management (TDM) strategies.

Travel models and associated post processing tools are used to develop and evaluate regional transportation plans, conduct regional air quality analysis required by FHWA, support NEPA planning studies for major transit and highway investments, corridor studies and planning, toll lane forecasting, Electric Vehicle (EV) infrastructure planning, Autonomous Vehicle planning, and Comprehensive Plans for counties and cities.