This chapter summarizes the panel's discussion on the topics of interest to BMC and the other meeting participants. Specifically, Section 6.1 includes responses to the questions the BMC posed to the panel at the outset of the meeting, and Section 6.2 addresses other meeting participant concerns and questions. For a list of all meeting participants, please see Appendix A.2
Question #1: How travel response to toll/pricing should be represented and modeled? Mode choice and route choice, or route choice only?
Panel's Recommendation: Mode choice components should be tied to high-level behavioral preferences (value-of-time (VOT) segmentation) and network components should be based on performance. Segmentation by value-of-time is a good idea. It is reasonable to do logit route-type choice, or alternately binary choice, within each segment. If the number of VOT segments mimics a continuous distribution, then binary choice will not be needed; if less, then still test binary choice.
Question #2: What spatial resolution should be used for discrete choice models? Parcel, micro-zone, or traffic analysis zone?
Panel's Recommendation: Use a consistent geographic micro-zone scale (e.g. parcels) across the entire BMC modeling region. Use empirical data to the extent possible and use disaggregation algorithms when not.
Question #3: At what spatial resolution should the Metropolitan Washington Council of Governments (MWCOG) juridsctions that are part of the BMC model area be modeled, given that there are inconsistencies between BRTB and MWCOG data?
Panel's Recommendation: See panel's recommendation for question two. Also, one of the audience members indicated that land use data for the MWCOG region is available at the parcel level.
Question #4: How should the transit modes be defined in the model?
Panel's Recommendation: Emphasize unique service variables in network coding (e.g. station type), elaborate path building choice logic, and keep the mode choice model shallow. Binary, or Conventional vs. Premium, distinguish between walk, P&R and K&R access/egress. Allow multi-class transit assignment by user class (income group and age).
In addition to providing recommendations to specific agency expressed issues, the panel also addressed the following audience concerns and questions since the discussion was open:
Question #1: Is 2007 household travel survey data too old to estimate the model?
Panel's Recommendation: No, year 2007 data is not too old. However, newer data will be useful for validating model sensitivity to background influences. Consider establishing a continuous survey program.
Question #2: What level of accuracy may be expected from the network assignment results?
Panel's Recommendation: ABM does not, by itself alone, guarantee better assignment results due to inherent limitations of static assignment and network accuracy. Effects of this project on goals of climate change will be limited to travel choice and constraints. The addition of more assignment time periods will assist in troubleshooting and re-calibrating network assignments. Many more solutions to assignment precision are available in dynamic traffic assignment (DTA).
Question #3: What are the dimensions along which the model should be validated?
Panel's Recommendation: Need to obtain diurnal traffic counts. There are many more dimensions for validation in ABM. Some can be internally validated against the survey. Consider validation of transit results per FTA guidance (e.g. district to district linked flows). It may not be possible to undertake certain types of disaggregate validation due to constraints of data availability (e.g. time-of-day, speed data).
Question #4: How is the effect of land use being incorporated in the model?
Panel's Recommendation: The current model structure seems to include land use effect only in the mode choice model. The entire model sequence should adequately reflect the interest in understanding the causal relationships between built environment, accessibility, and travel choice (i.e. beyond mode choice).
Finally, the panel made the following additional suggestions/observations that may provide useful guidance during the model development process:
In June 2013, BMC started the process of developing an activity-based model (ABM) for the Baltimore region with a goal to have a fully functional model by June 2016. In December 2013, the agency organized the current peer review to seek guidance and recommendations on a number of key issues that may enhance the model development process, including design framework, methodology to address policies of interest, model validation, and sensitivity testing. Next, the agency and the consultant will review the panel's recommendations and identify the most effective ways to incorporate the findings as they continue with the ABM model development effort. Once substantial progess has been made, but prior to model implementation phase, the agency intends to seek guidance from the panel again on issues such as model structure, estimated parameters, elasticity, and performance.
