Currently Adopted Travel Model

The TPB travel forecasting model serves a number of functions. It provides regional travel demand forecasts and air quality assessments to support both long-range transportation planning and the development of the six-year Transportation Improvement Program (TIP). It supports ongoing corridor and sub-area studies, which are typically performed for member state and local governments with consultant assistance. The model is also used to examine the mobility of various population segments.

The TPB travel model is periodically refined with newly collected data or with emerging forecasting techniques. At any point in time, there will always be an "official, adopted" travel model of record reflecting the last approval action by the TPB. The Version 2.3 Travel Model became the adopted regional travel model for the for the Metropolitan Washington Region on November 16, 2011, when the TPB approved the Air Quality Conformity Determination of the 2011 Financially Constrained Long-Range Transportation Plan (CLRP), since that was the model version used for the analysis. Prior to that approval, the Version 2.2 Travel Model was the adopted travel model. This model is the result of a multi-year "models development" program, which is conducted by TPB staff, often with consultant assistance, under the oversight of the Travel Forecasting Subcommittee, which reports to the TPB Technical Committee. The Version 2.3 Travel Model is described below.

Model Documentation

The TPB Version 2.3 Travel Model is documented via a calibration report, a validation memo, and a two-volume user’s guide, which can be found on the Model Documentation page:

Model Description

The following lists some of the key distinguishing features of the TPB Version 2.3 Travel Model:

  • A trip-based, "four-step" travel model, calibrated and validated with a number of recent data sets, including:
    • 2007/2008 TPB Household Travel Survey
    • Transit on-board surveys
      • 2008 Metrorail Survey
      • 2008 Regional Bus Survey
      • 2007-2008 On-Board Survey of Maryland Transit Administration (MTA) Riders, which would include users of MARC train service
      • 2005 Virginia Railway Express (VRE) Passenger Survey
    • Recent traffic counts
  • Modeled area
    • Covers 22 jurisdictions and about 6,800 square miles
    • Includes the District of Columbia and parts of three states: Maryland, Virginia, and West Virginia
  • A new transportation analysis zone (TAZ) system containing 3,722 TAZs (this is almost double the number of TAZs used in the previous version of the travel model).
  • Transportation network
    • Highway network: Contains about 45,000 links, of which about 30,000 represent non-centroid links
    • Transit network: Contains about 1,000 transit routes for the AM peak period
  • Trip generation for five primary trip purposes: Home-Based Work (HBW), Home-Based Shop (HBS), Home-Based Other (HBO), Non-Home-Based Work (NHW), and Non-Home-Based Other (NHO).
  • Three other trip purposes:
    • Non-freight, commercial vehicles (including autos, vans, and light duty trucks)
    • Medium trucks (two axles and 6 tires)
    • Heavy trucks (all combination vehicles)
  • Trip generation of both motorized person trips (person trips in cars, buses, and trains) and non-motorized person trips (walk and bike). Only motorized person trips continue through the model to trip distribution, mode choice, and trip assignment.
  • Trip distribution model uses the standard gravity model formulation and makes use of a composite-time function that represents a blending of transit and highway travel times. The gravity model is doubly constrained for all five trip purposes.
  • Mode choice model:
    • One for each of the five trip purposes
    • Nested-logic mode choice model with 15 travel modes, including
      • Three auto modes (drive alone, shared ride 2-person, and shared ride 3+person)
      • Four transit modes (commuter rail, all bus, all Metrorail, and combined bus/Metrorail)
      • Three modes of access to transit (park and ride, kiss and ride, and walk)
    • Although the nesting structure does not include explicit branches for specialized transit modes - such as light-rail transit (LRT), bus rapid transit (BRT), and streetcar - the model is designed to deal with these special transit modes.
  • Time-of-day model apportions daily resident travel among four time periods:
    • AM peak period (6:00 AM to 9:00 AM),
    • Midday (9:00 AM to 3:00 PM),
    • PM peak period (3:00 PM to 7:00 PM), and
    • Nighttime/early morning hours (7:00 PM to 6:00 AM).
  • Traffic assignment
    • Static, user equilibrium assignment
    • Six user classes arranged across six traffic assignments
    • Stopping criterion: A “progressive” relative gap threshold, which varies from a value of 0.01 (10^-2) in the early speed feedback iterations to a value of 0.0001 (10^-4) in the final (iteration 4) speed feedback iteration
  • Equilibrium of model results
    • Speed feedback loop recycles constrained traffic speeds from highway assignment back into earlier modeling steps
      • Peak-period transit accessibility measures fed back to demographic sub-models
      • Peak and off-peak composite (highway/transit) travel times fed back to trip distribution.
      • Five executions of the travel model for each alternative tested, via five iterations of the “speed feedback” process (one initial or “pump prime” execution of the model followed by four normal executions of the model).
  • Updated truck models
  • Incorporation of parallelization/distributed processing, via both native Windows techniques and Cube Cluster’s intra-step distributed processing (IDP) and multi-step distributed processing (MDP). This allows two or more parts of the model to run simultaneously, helping to minimize model run time.

Click here for information on how to request models and data.