With the 12^th highest regional gross domestic product (GDP) in the U.S.[1], Southeast Florida is a major economic engine. Transportation infrastructure and services supports this economy, underlying the movement of people and goods throughout the region. And, the Southeast Florida Regional Planning Model Version 8 (SERPM 8) is a critical tool in identifying and evaluating appropriate investments to develop and maintain this region’s transportation system.

SERPM 8 is the travel demand model calibrated to a 2015 base year and 2045 forecast year for southeast Florida covering Palm Beach, Broward, and Miami-Dade counties.

Features of this Model

SERPM 8 is an activity-based model. This framework was adopted for Southeast Florida beginning with SERPM 7 and the SERPM 8 overall model structure is the same as the SERPM 7 model structure. This type of model has become the state of the practice in travel demand forecasting in the largest U.S. metropolitan areas. This type of model is preferred in such areas because it is grounded in an explainable modeling framework that simulates how discrete travel decisions are made, as opposed to being only a mathematical construct reflecting aggregate regional characteristics of travel. The behavioral grounding and disaggregate nature of the activity-based modeling framework allows a variety of additional questions to be tested and for additional detail to be available to decision-makers working with the model output.

The chief emphases of the SERPM 8 model update were:

· update the model base year from 2010 to 2015 (with the related changes in the socioeconomic data and transportation network inputs);

· simplify the mode choice model structure to allow for dynamic transit service definitions and testing and improve model validation;

· update model parameters to reflect new information gleaned from the regional household travel survey which preceded; and

· improve model usability and training aids to support a broader range and more capable set of users in the region.

Key Model Components

The resident travel portion of the model applies a sequence that considers long-term choices of school and work location for students and workers, followed by household-level choices on car ownership, technology, and transportation network company (TNC – e.g. Uber, Lyft) membership. Next, for each person in the synthetic population, the level, location and timing of mandatory and non-mandatory travel are simulated. Coordinated travel within a household, fully-joint travel, is also simulated. Once the number, location and timing of tours is simulated, the tour-level choices are made around mode and, if there are intermediate stops, the stop location and timing, followed by trip mode. Finally, trip assignment steps are undertaken to determine specific roadway and transit service usage. Visitor travel internal to the region is also simulated disaggregately although with fewer components and segments than the resident models.

The model also includes treatment of trips by air passengers, cruise passengers, external travel, and truck travel through model frameworks that are based on mathematical constructs reflecting aggregate regional characteristics of these forms of travel. In general, changes in these types of travel are not the primary focus of the long-range investment decisions that are the primary use-cases of the regional travel demand forecasting model and therefore it is reasonable that simplified approaches are employed (as compared with the activity-based resident and tour-based visitor travel modeling).

Future Mobility Support

In most cases, future scenarios can be represented by the core activity-based model component sensitivities. For example, changing land use scenarios, demographic shifts, and changes in the roadway and transit networks are all supported by the ABM components estimated from household surveys and calibrated to base year conditions. However, representing new vehicle technology and mobility services require an extension of the core activity-based model sensitivities. SERPM 8 includes several components and enhancements to support the testing of "Future Mobility" scenarios. The types of future mobility supported by these components are:

• Connected / self-driving / driver-less vehicle impacts on behavior and network performance;

• Mobility services provided by TNCs; and

• Vehicle technology threshold for managed lane usage.

Cube Flowchart

The overall model structure is reflected in the SERPM 8 Cube Flow Chart interface, shown below.

The general model process is:

• Conduct one-time preprocessing of highway and transit networks (Initialize Inputs and networks)

• Conduct one-time generation of non-activity-based model components (Non-ABM Generation)

• The speed feedback loop to converge travel demand with supply conditions, this loop consists of:

  • Update networks with latest conditions

  • Estimate resident and visitor internal travel demand (ABM Components)

  • Estimate truck, special generator (airport and cruiseport), and internal-external trip distributions (Non-ABM Distribution)

  • Load newly estimated demands onto the highway network (HWY Assignment)

• After convergence, the Transit network is loaded (Transit Assignment) and reports are generated (Post Processors)