What we are trying to do [and why]


  • Persistent struggles with transit service quality and overcrowding

    For more than a decade, many urban transit systems have struggled to meet their service standards for both reliability and capacity provision. For example, San Francisco's Muni system is quite crowded, with official reports showing that over 5% of peak period travel happens on transit lines that are over 125% of their peak load capacity. The passenger experience is further degraded by the unevenness of the vehicle arrivals as only 6 out of 10 arrive within a few minutes of their scheduled headway . As a result, people are turned away from traveling on transit and on to congested roadways both because they cannot physically get on and because it can not reliably get them to work, appointments, or home.

  • 2010/2013

    Plans adopted - without robust reliability or capacity analysis

    While qualitative emphasis was placed on transit capacity and reliability for the adopted regional transportation plans in both the Bay Area and Puget Sound regions, no quantitative analysis of the repercussions of the "failure to act" were included because proper tools and data did not exist.

  • 2014

    Grants won for more robust transit capacity and reliability tools development and analysis

    The historical emphasis of studying and planning-for vehicle-travel rather than person-travel has resulted in a multitude of heavily-researched and widely-accepted technical tools to support vehicle-travel, but few if any to analyze transit reliability and capacity. Following the adoption of the San Francisco Transportation Plan and Plan Bay Area, planners and policy makers applied for and won two significant grants to address this disparity ahead of the next round of regional plans. The Metropolitan Transportation Commission led two multi-agency collaborations:

  • Winter 2015

    Project Kick-Off

    Following various contracting and startup steps, we officially kicked off the project in winter 2015

  • Data Standards and Methodology Development

    The first major step is to complete background research on related methods and practices in order decide on the best methodological approach for our own work. At the same time, we will need to develop data standards to be used for our study, because existing standards such as GTFS do not include all of the fields and information that we need in order to capture relevant transit features.

  • Data and Software Development

    Over the summer, the Fast-Trips software will be refactored form C++ to Python. Additionally, four pieces of data will be developed and refined:

    1. Transit network data for the Bay Area and Puget Sound, which will be based on an expanded GTFS data standard and work with existing travel model specifications
    2. Transit demand data by market segment will be based on the data from each travel model, but then calibrated to observed demand based on transit on-board surveys, fare card data, and automatic passenger counter data. While in the final system the demand will be endogenous, this observed demand will be necessary to ensuring the model is appropriately calibrated and validated to existing conditions
    3. Transit route data by market segment will be processed from GPS data to match the transit network data and used to estimate the route choice models in the fall
    4. Transit performance data such as travel times and reliability will be processed from raw automatic passenger counter and vehicle location data in order to compare the Fast-Trips simulation to reality
  • Software Testing and Transit Route Choice Model Estimation

    In addition to using this time to improve the performance of the Fast-Trips transit simulator, we will be estimating route-choice models from observed transit route data using BIOGEME software in order to deduce how much transit passengers of various market segments (i.e. youth, commuters, elderly etc.) value various service features (time, reliability, seat availability, walking distance etc.) of each component (walk access, waiting, transfers, in-vehicle time) of their transit route. These parameters will be used within the Fast-Trips simulation.

  • Travel Model Implementation

    This is "putting it all together" with respect to inserting the new transit simulator, Fast-Trips, into each agency's existing travel forecasting tool:

  • Calibration, Testing, Documentation, Education

    The final step involves running the entire transit model system and making sure it is a sufficient representation of reality in terms of overall accuracy of forecasts, but more importantly that it exhibits appropriate sensitivity to service quality and capacity changes so that it can be used in evaluating planning and policy changes.