Following the presentations by SEMCOG staff and their consultants, the panel convened to discuss and develop their recommendations. These recommendations were presented to the attendees at the peer review meeting and are described in this section of the report. The peer review panel grouped their recommendations into six main topics, and for each of these topics provided a topic definition, planning objective, problem statement, basic solution, advanced solution, and resources. The basic solutions are intended to be short term recommendations that are applicable to SEMCOG's current model development program, while the advanced solutions are intended to be long term recommendations that are intended to inform SEMCOG's planning for future model enhancements. The six topics are as follows and are discussed in turn below:
Household surveys are becoming increasingly difficult. For example, the usefulness of telephone surveys ended when wireless telephones became ubiquitous. Therefore, it is clear that a new paradigm is required.
The American Community Survey administered by the Census Bureau is an example of a continuous survey that is easier to budget for and provides the opportunity to correct and improve the survey incrementally. However, it is becoming clear that the expected cost efficiency of this type of survey is not materializing. Instead, new approaches such as using probe data might become the state of the practice. Sampling is also a growing issue, with non-response bias a concern. This suggests the need to focus more efforts on hard to reach populations and then deal with the subsequent weighting issues.
With respect to the development of activity based models, it will be important to focus on just the data items that are required to support those models, and use more innovative survey forms such as stated preference surveys to help understand travel behavior.
It is important to distinguish between external travel surveys and long distance travel surveys. The panel recommends not doing long distance surveys as part of a household survey as the level of success of these types of surveys is low. Instead targeted surveys at borders, hotels, and convention centers should be considered.
Freight model data needs are growing but it is important to recognize the difference between truck trip table development and understanding the bigger picture. Understanding the bigger picture, where truck travel is derived from freight movement and the economy means that some understanding and linkages with economic modeling and planning are required. Collecting data in this area is likely to require infiltrating the freight industry to investigate supply chains and answer questions about how commodity movement decisions are made, how industry transformations will affect truck trips, and how through movements such as intermodal container transfers operate.
The panel recognized that SEMCOG has very good highway network performance data. They recommended that SEMCOG investigate partnering with GM OnStar as a potential source for probe data on speeds and reliability. GM OnStar might be interested in sharing their data for the public good in their home city.
To support equity analysis, the panel recommended careful consideration of special surveys for special needs. For example, transit modeling is typically focused on the trade-off between transit and auto for those who have that choice and less on the mobility benefits for people who have no choice but to use transit. Adequately representing the travel choices of this group might mean understanding issues such as the effects of job-housing balance, the locations of affordable housing, undocumented residents, and the underground economy.
The panel recognized that static assignment is adequate for planning capacity additions but it is not adequate for measuring system efficiency. For that, DTA is the tool but it is still experimental. Furthermore, for the benefits of DTA to be fully realized it needs to be coupled to an activity based demand model. Coupling a DTA to a trip-based model with temporal and spatial disaggregation will achieve some of the benefits. DTA packages are not off the shelf yet and so the panel recommends that SEMCOG consider developing internal skills to support implementing a DTA. However, DTA is still the province of consultants working on SHRP, NHCRP and FHWA funded research projects, and universities.
The panel confirmed that their recommendations for operations modeling are in line with the recommendations to transition from SEMCOG's E series models to the A series models made by Parsons Brinckerhoff.
Freight modeling is a growth area for travel modeling: the current practice is wrong, but more advanced methods are currently undeveloped. The status of freight modeling in the Chicago region was discussed by the panel. CMAP (the Chicago MPO) is moving towards implementation of the framework that was initially developed by LA Metro. This was partially implemented as CMAP mesoscale freight model, and is now moving forward with the demonstration of a combined supply chain and truck touring model, supported by an FHWA funded project that is nearing completion.
The panel discussed the conflicting needs for transit modeling in the SEMCOG region, including maintaining a complex transit model to represent a system with low ridership, supporting the very specific requirements of preparing New Starts forecasts, and how transit modeling will fit into the advanced modeling framework being considered.
The panel recommended that, if the transit modeling needs are limited to smaller incremental changes to the transit system, then simple mode choice approaches, such as a pivot analysis from a trip table of existing transit riders, should be used. The most important variables describing likelihood to use transit might be auto availability and age. In general, only the travelers with auto availability can respond to relative levels of service represented by variables such as in-vehicle time, wait and fare. Riders without an auto available might respond to changes in transit routes by walking to another line, changing to bicycle, finding a ride from someone, or not traveling.
However, as the transition to more advanced models is made and more significant transit improvements are considered, in order to predict behavior accurately it is very important to identify travel market segments more fully and develop models that represent those markets in the AB model context. This means that they must be represented in the population synthesis and in the travel microsimulation. For example, the physically handicapped population might be much more likely than the able bodied population to use transit. To represent this in the travel model it is necessary to survey them, model them, and represent them in the population synthesis and all subsequent stages in the microsimulation.
Traditional four step models use economic and land use inputs as fixed inputs, but this is now recognized as being wrong. The panel recommended starting with SEMCOG's existing UrbanSim model and experimenting with feeding back accessibility measures from the TDFM. The panel suggested more advanced modeling would involve use of an economic and land use simulation that is sensitive to declines and shifts in economic relationships; however, they recognized that there are no tools available that deal with permanent or sustained regional economic decline. Oregon provides a good example of a possible approach for Michigan and the SEMCOG region, with a PECAS model implemented at the statewide level that is used for economic forecasting, and then metropolitan land use models that are used for land use allocation.
The panel explained that in order to adequately model equity impacts, microsimulating people is key, which is a feature of AB models. The aspects that must be represented in the model include the connection of people to land use; for example, correct representation of demographics and the connection of people to work and school, and how these relationships might change over time as land uses change.