Southeast Florida Transportation Council (SEFTC) Peer Review

3.0 Development of the SERPM 7.0 Model

3.1 Existing Model

This section presents a summary of the existing SERPM 7.0 model. The information provided here is drawn from the draft report Southeast Florida Regional Planning Model -- SERPM 7.0, prepared by Parsons Brinckerhoff Inc. (PB), The Corradino Group, Inc., and BCC Engineering, Inc., dated February 2015.

SERPM 7.0 comprises several model components that address the vast majority of urban travel in the SEFTC region. At its core is the internal resident travel model, which is an activity based model (ABM) implemented using the CT‐RAMP framework. This structure is the basis for forecasted travel for residents that start and end within the region, with the exception of travel to special events. Travel generated by overnight visitors is forecasted by the Visitor Model, which is also based on an ABM framework. External‐external and internal‐external travel is forecasted by traditional trip‐based model components. Finally, like its predecessors, SERPM 7.0 includes a truck model and an airport ground access travel models. At this time, SERPM 7.0 does not model travel to special events.

The CT‐RAMP framework, which is fully described in the following section, adheres to the following basic principles:

• Models individual and joint travel choices with a high degree of behavioral realism. In particular, it addresses both household‐level and person‐level travel choices including intra-household interactions between household members, and employs disaggregate (i.e., individual decision‐maker) models throughout the model system.
• Operates at a detailed temporal (half‐hourly) level, and considers congestion and pricing effects on time‐of‐day and peak spreading.
• Reflects and responds to detailed demographic information, including household structure, aging, changes in income, and other key attributes.
• Implements the Common Modeling Framework, an open‐source library developed by Parsons Brinckerhoff specifically for implementing advanced travel demand models.
• Offers sensitivity to demographic and socio‐economic changes observed or expected in the dynamic SE Florida metropolitan region. This sensitivity is ensured by the enhanced and flexible population synthesis procedures as well as by the fine level of model segmentation. In particular, SERPM 7.0 incorporates different household, family and housing types including a detail analysis of different household compositions in their relation to activity‐travel patterns.
• Accounts for the full set of travel modes. PB's experience with previously developed ABMs has shown that mode choice is one of the least transferable model components, because each region has a specific mix of modes developed in the context of the regional urban conditions.
• Integrates with other model components. The CT‐RAMP model is one component (person travel) and can integrate with other components such as the existing SERPM truck, airport and external models.

SERPM 7.0 was developed by transferring ABM components for resident and visitor travel developed for the San Diego Association of Governments (SANDAG) and adapting the trip‐based model components of previous versions of SERPM. The SANDAG implementation structure was chosen over other CT‐RAMP options primarily due to its treatment of transit access and similarity of transit options ‐ the spatial representation of home, stops and activity locations is based on micro‐area zones, which improves the calculation of walk access/egress distances, and the set of current and planned transit modal options is the same between the two regions, which greatly simplifies the model transfer. The majority of network‐based procedures, such as highway and transit skimming and assignment were adapted from earlier SERPM versions (6.5 and/or 6.7).

The 2009 National Household Transportation Survey (NHTS) add‐on was used intensively in the development of SERPM 7.0. While the sample is insufficiently large to support the original estimation of most of the submodels that comprise SERPM 7.0, it provided sufficient information to develop region‐wide calibration targets for most submodels. In developing these calibration targets, the NHTS was supplemented with a wide variety of other data sources, including Census and American Community Survey (ACS) data, Longitudinal Employment‐Household Dynamics data, data from the Florida Department of Motor Vehicles, SunPass account sales data, transit on‐board survey data, transit ridership data, and school attendance data, among others.

The SERPM 7.0 calibration targets were compared to similar targets developed for other regions, to verify that the aggregate tabulations of travel behavior across various person types and types of travel exhibited similar relationships. Because the SERPM 7.0 submodels could not be estimated with local data, a complete model specification was adopted, patterned after SANDAG's ABM deployment. The adequacy of the model transfer was evaluated by examining how well the transferred model, without updates, matched the calibration targets developed from the NHTS. The transferred model was in fact able to reproduce fairly well the Southeast Florida travel behavior at an aggregate level. The submodels that performed least well are, not surprisingly, the tour and trip location models. This can be explained partly by differences in model region size between San Diego County and the SEFTC region, and partly due to differences in multimodal accessibilities and the composition and location of employment. The development of the tour‐level, mode choice submodels also relied on various relationships of transit tours to transit trips obtained from a recent Atlanta onboard passenger survey, given the near lack of transit tour observations in the 2009 NHTS sample

3.2 SEFTC's Goals for the Current Peer Review

Prior to meeting, SEFTC identified two main areas for which they wanted the peer review panel to comment and make recommendations. The two main areas and related questions are detailed below:

1. Guidance on best-practice approaches to assess SERPM 7.0's performance, given that the model parameters for the ABM and related sub‐models are borrowed from another region with constants that were calibrated so the model reproduces aggregate targets developed from local data.
   • Is the magnitude of (and change in) the constants reasonable?
   • Does the calibrated model perform well, globally and by submodel/subarea/mode?
   • Are there additional tests that need to be performed or presented to better assess its performance?
   • Are there best‐practice benchmarks that the model can be compared against, for example to establish acceptable ranges for trip/tour frequency, trip distance, mode share, etc., elasticities with respect to travel time and travel cost?
2. Guidance on data collection effort to undertake in the short term (next 2 years), medium term (2‐4 years), and long term. The proposed data collection effort should be based on the performance and structure of SERPM 7.0, and consider all of the major travel markets that the model is expected to address: residential travel, overnight/visitor/seasonal resident travel and truck travel.
   • With what frequency should the various datasets be collected?
   • Should the next phase of the data collection program be focused on data required to re‐estimate some or all model components, data required to calibrate/validate the model outputs, and/or data required to verify the model inputs?
   • What readily available "off the shelf" datasets have the panel found most useful/reliable in their efforts and which datasets are best collected locally?