When VTrans presented the questions above to the peer review panel in Sessions #1 and #2, the panel was able to ask questions to gain further detail with respect to each question and provide initial feedback. The panel then convened without VTrans staff present in Session #3 to further assess the appropriate responses to each question based on the experiences of each expert panelist. The following section details the responses provided by the panelists as presented to VTrans in Session #4.
The panelists recommended that VTrans address each of the fundamental model development considerations provided from FHWA. Particularly, the panelists agreed that the development of both a users' guide for staff training purposes and a detailed technical reference document containing codes, parameters, assumptions, and other intricate model specifics would be helpful for communication purposes.
The panelists also stressed the importance of defining short-, mid-, and long-term priorities based on the status of the existing statewide model. The panelists noted the ambitiousness of VTrans in their exploration of a wide myriad of application possibilities for the statewide model; however, given logistics, schedule, and budget, prioritization of these enhancements will be critical to creating an effective model development plan.
For example, the panel suggested that in the short-term, it would be valuable for VTrans to develop one comprehensive statewide model package, assess the consistency between sub-models, and establish the appropriateness of model to meet agency needs. In the mid-term, the panelists recommended that VTrans identify a second phase of model enhancement based on the agency's next strongest priorities. These model enhancements could then occur while the model package developed in the short-term enhancement phase is in application. Finally, considering long-term model enhancement the panel recommended that VTrans compile a "wish list" of features that are lacking from the short- and mid-term enhancements, prioritize these features, and work towards them incrementally based on agency needs once the second phase of model improvements is complete.
The panelists derived a variety of potential options for the assessment of infrastructure design. One panelist initially noted that a travel model will establish a state of equilibrium in which long-term travel decisions are reported, rendering the model insensitive to disaster response. The assessment of infrastructure design and ability to withstand extreme weather or environmental circumstances will require structural design specification evaluation combined with spatial analysis of environmental factors, studies for which a travel model is not applicable.
The panel also noted that emergency contingency planning is associated with links damaged by an emergency event, not general facility design. In this sense, emergency contingency planning for road closures is an appropriate use of the statewide model for particular analyses. It was suggested that VTrans consider the use of a dynamic traffic assignment methodology to assess traffic patterns in emergency response situations or structural failures.
In the event that VTrans would like to apply the statewide model for emergency scenario planning, the panelists recommended that VTrans staff identify the criteria for use in the comparison and evaluation of alternative scenarios. These established performance measures could then guide the development of model features.
To apply the statewide model for emergency scenario planning, VTrans would need to develop an at-risk location inventory in the highway network through the addition of link attributes, marking links with a potential for failure. For example, the model network could be set up to incorporate geo-coded bridge inventory data in order to "fail" bridges and mimic emergency situations. The model could be run with and without failures, and the impacts to traffic flows could be compared between scenarios. Additionally, it would be helpful to design the network to include references that enable the incorporation of agency data into the model network in an automated manner.
The panelists initially identified the necessity for a mode choice component in the statewide model in response to VTrans's interest in using the statewide model for energy use and emissions evaluation. The model will require the design, estimation, calibration, and integration of SOV, HOV2, HOV3, air, rail, bus transit, bike, and walk mode choice models. Additionally, sufficient travel survey information will be required to accurately develop these mode choice models. Furthermore, the model network and TAZ structure will necessitate further resolution to accommodate for bicycle and walk trips, while the model network will need to be expanded to represent rail and air service. The model's sensitivity with regard to parking pricing, transit pricing, toll pricing, and other types of operational costs will also need to be refined to allow for mode shifts based on deltas in price, travel times, land use inputs, and other factors of travel choice significance.
When the model has the desired modes represented and is appropriately sensitive to mode choice factors, it can be used to produce trips by mode, trip length by purpose and mode, and VMT by vehicle classification. Emissions can then be estimated by applying this output in combination with the Motor Vehicle Emission Simulator (MOVES).
While the incorporation of mode choice, specifically rail and transit, in the statewide model would be beneficial, the panel highlighted the difficulty in addressing the performance measures needed for emissions assessment given the scale and resolution of the statewide model. The panel noted that energy use and emissions analysis may be more appropriately conducted using MOVES coupled with a regional model that has an existing detailed TAZ and network structure, particularly for non-motorized travel estimation. Ultimately, an activity/tour-based model would be required to yield meaningful results for emissions assessment rather than application of derived or assumed trip tables.
Another strategy for emissions analysis suggested by the panel was the development of a separate aggregate model that combines information from the statewide model and the Chittenden County MPO model to evaluate VMT, transit/rail trips, auto occupancy, and non-motorized trips. Specialized tools, such as an aggregate model, often entail fewer resources, yet can provide required elements for agency reporting while taking advantage of model improvements made to either the MPO model or the statewide model.
The panel came to a general consensus that staging the model's enhancements should be primarily premised on basic functionality. The qualities listed above that are required for energy and emissions scenario evaluation could be incorporated in the long-term, while model enhancements to improve basic functionality should be prioritized in the short-term.
The panel's recommendation in response to VTrans's ambition to conduct system preservation and disinvestment analyses using the statewide model was to first identify the performance measures desired for project prioritization before proposing enhancements or adjustments to the model. These performance measures will be used to evaluate different scenarios and test various combinations of projects over time to reveal the outcomes of each set of projects; therefore, the performance measures selected by VTrans will be imperative in determining the staging of model enhancements. The panel suggested that VTrans consider using the statewide model to evaluate volumes and road wear as performance measures for project prioritization
For an example of an existing methodology with a successful record in optimizing transportation investments, the panelists suggested that VTrans review Oregon's use of the HERS-ST model. Growth rates from the statewide model can be applied to the HERS-ST model applies to evaluate forecast needs and options for transportation investment optimization. It was also noted that Caltrans conducted a study to evaluate the effect of road conditions on truck speeds, maintenance costs, and costs associate with damaged goods, which could be further referenced.
The panel noted that because Vermont's Highway Performance Monitoring System (HPMS) currently reports pavement conditions and the State Highway Operation and Protection Program (SHOPP) prioritizes system preservation projects, it was recommended that these groups along with VTrans pavement program staff are contacted prior to model enhancement for system preservation and disinvestment evaluation purposes. These agencies may have sufficient evaluation methodologies in place or find minimal need for this level of effort.
In the discussion of model application for performance-based planning and asset management, the panelists again placed an emphasis on the importance of (1) determining which performance metrics are critical with regard to agency needs, (2) establishing the model improvements needed in order to calculate these improvements, and (3) creating a prioritized list of these metrics and associated improvements to guide the model development plan. Given the model development plan design, the timing at which specific measures can be reported will depend on the model development phase, reiterating the importance that the phasing of model development features be dependent on agency priorities.
The panel acknowledged that the statewide model is the only tool available to VTrans to forecast MAP-21 criteria associated with vehicular delay and system reliability. Credible system utilization forecasts, as they relate to factors that affect system conditions (e.g., volumes, classifications, etc.), can provide valuable information to decision-makers to help anticipate facility life-cycle impacts. Additionally, economic assessment software (i.e. - STEAM, T-PICS, TREDIS, REMI Transight) may be applied to the travel model output to evaluate performance measures such as job creation and gross state product (GSP).
Additionally, the panel suggested that a separate post processing methodology could be developed to determine the economic impacts and GSP values of individual links. This component could be developed in the mid- to long-term.
As an alternative to separate economic assessment software or application of a post processor, the panel recommended that VTrans consider employing a separate project-specific benefit/cost model. One panelist noted that Caltrans is in the process of evaluating model performance measures for the next California Transportation Plan and MAP-21 efforts. There are the "standard" outputs of VMT, mode split, trip length distribution, emissions, vehicle hours of delay/travel, and congested speed that can each be used as part of the MAP-21 related efforts; however, the agency does not anticipate performing model upgrades as a specific response to MAP-21. Caltrans uses a separate benefit/cost model (Cal B/C) that is project specific. This specific methodology should be reviewed by VTrans.
The panel sited that impact fee policy scenarios could be assessed using the statewide model to develop estimates of VMT for new development by land use type and trip purpose and determine change over time. The panel, however, suggested that statewide model resolution is not adequate for the post processing methodology that would be necessary to determine long-range growth rates for background traffic. The resources required for this type of spatial accuracy would far exceed the analytical gain from this level of detail. Therefore, it was recommended that VTrans consider using a micro-simulation model, which applies future volumes and growth rates from the regional model, for analysis of specific improvements to meet demand associated with new developments.
It was also recommended that VTrans explore the possibility of using off-model techniques for development impact assessments. For example, the ITE Trip Generation Manual details data available for various off-model methodologies that can be applied in these types of studies separately or in conjunction with the travel model.
The panelists found it important that VTrans recognize that the statewide model may not be the appropriate resolution for evaluating non-motorized transportation improvements. A parcel level TAZ structure and additional travel survey data regarding current travel by mode and future preferences would be required for this type of analysis. The panel noted that a statewide model can effectively handle air, rail, and bus transit, truck, and car modes properly granted the scale of the improvement is significant enough, the TAZ structure is small enough, and the road network is detailed enough so that all travel is not deemed intra-zonal and travel times are not significantly skewed.
The panel recommended that VTrans develop a tool separate from the statewide model for application in the estimation of non-motorized transportation. While not conjoined with the model, this tool should still be compatible and consistent with the statewide model. The panel suggested that VTrans consider the implementation of a micro-simulation model for local area analysis. It was recommended that VTrans also consider a survey effort to further the agency's understanding of current travel by mode within the state.
The panel noted that regional models that perform well in evaluating non-motorized alternatives typically require advanced techniques, such as activity-based modeling, to better address non-motorized alternatives. If VTrans envisions non-motorized travel modeling as a major priority for the statewide model, it is possible that the agency could transition to an activity/tour based model to increase sensitivity to the drivers for bike and walk mode. For this transition to occur, it was recommended that VTrans consider a tiered approach to activity-based model development for non-motorized travel to gradually increase model resolution as a long-term priority.
Based on the background information presented by VTrans and the discussions in the first two peer review sessions, the panel identified issues in the existing model. These were primarily based in the model network. Other potential issues hinging on the model's structure were also identified.
The panelists advised that the network model area be extended beyond the political boundary of the state with, at a minimum, a halo region around the state to represent activity related to Vermont's adjacent states. Because Vermont is such a small state, a significant amount of Vermont's travel crosses the state line, making the development of a robust external model for this state critical. The panelists also identified the reassessment of network coding as a necessity for the model. Centroid connectors, link attributes, and potential missing roads should all be comprehensively checked for quality assurance.
Other structural elements of the statewide model requiring attention that were identified in the panelist discussion were: