Recommendations from the peer review panel are divided into two categories:general comments and response to technical questions.A summary of recommendations from the panelists follows.
Members of the panel agreed to several general comments and recommendations that weren't necessarily in response to specific questions. Those comments are summarized in this section.
Overall, the CHCNGA-TPO has done a good job in expanding its resources and building a solid foundation for meeting short, medium, and long term goals.This is particularly impressive given the relatively modest size of the region.
The CHCNGA-TPO has done a good job in data collection and development of data sources; however, it is important to realize that model development efforts shouldn't be driven by the availability of data. Similarly, when considering model development, it is important to match the expectations of the development task with the commitment of resources.
Being regional in nature, the travel demand model is probably not the best tool for modeling non-motorized travel.Instead, specific non-motorized modeling tasks should be performed at a corridor / sub-area level using a tool designed for such a use.
As the CHCNGA-TPO considers adding a transit mode choice component to the travel model, a relatively simple multi-nomial logit model is recommended.This type of model should meet the needs of the CHCNGA-TPO without being too difficult or time consuming to develop.
In developing a land use model, it is recommended that the CHCNGA-TPO move forward with the CommunityViz model. CommunityViz can be developed more quickly than UrbanSim, which will make it easier for the CHCNGA-TPO to maintain their overall modeling schedule.However, this doesn't mean that consideration of UrbanSim needs to be abandoned.It is recommended that UrbanSim be the subject of further research and possible future implementation.This research should include talking with current UrbanSim users to learn about their experience.Additionally, the CHCNGA-TPO may consider partnering with a university to share the development load.
Regarding CommunityViz, it is important to remember that it is a land use allocation model that requires control total information as an input.The CHCNGA-TPO should investigate if there is a regional or statewide REMI or other economic model that can provide assistance in developing these control totals.
On the subject of the highway network, the CHCNGA-TPO explained that the model network was originally based on TIGER files and has subsequently been cleaned and is compatible with the region's GIS street centerline file.The panel felt that the CHCNGA-TPO should consider refreshing the network and utilizing a NAVTEQ or other similar street centerline file.
The CHCNGA-TPO stated that they would like to commence with modeling to support the next Long Range Transportation Plan in 2012.With this deadline in mind, it is essential that work begin as soon as possible on the model update tasks (including the land use model) so that they are ready for use in time.
On the subject of balancing the model development workload between CHCNGA-TPO staff and consultants, the panel felt that staff's involvement with a land use model is more important than with the travel model.This is because of the need for local knowledge and relationships in developing a land use model.Nevertheless, it is important for staff to stay involved with the travel model development process to ensure ownership of the process and the results.
In managing staff resources, the panel thought that workload be considered as part of the "production track" or the "research track."The production track would focus on the tasks necessary to producing the CHCNGA-TPO deliverables, which could include short-term model development efforts.The research track would focus on medium- and long-term projects.Adequate time would need to be given to each track for the CHCNGA-TPO to achieve its goals.
This section describes the peer review panel's response to specific technical questions.
With our limited resources and equipment, how can we improve our non-motorized data collection process in a manner that will support travel demand modeling?
As mentioned in the Recommendations section, the panel felt that the CHCNGA-TPO should consider off-model tools to analyze non-motorized travel for specific corridors or study areas. It was felt that the near and medium term benefits associated with modeling non-motorized travel would not be commensurate with the effort necessary to achieve such a goal.It was suggested that non-motorized trips could be removed from the model after trip generation so that they would still be included in the total trips in the region.It was also suggested that non-motorized travel after trip generation could focus on transit access modes rather than as a stand-alone mode.
Based on the various data sources available, what is the best source and the best methodology for estimating and forecasting employment?
The panel felt the best employment data source to be the Quarterly Census of Employment and Wages.This data could be cross referenced with field surveys, the phone book, and commercial data.The CHCNGA-TPO should consider use of a REMI or other economic model to provide control total information (probably in coordination with the state or other MPOs).An expert panel could also be used to validate employment data.
In developing employment data, it is important to consider how the data will be used in developing employment categories.There is no need to develop data for numerous categories if they are simply going to be aggregated before use in the model.This approach will enable the CHCNGA-TPO to focus their data collection on what will actually be used by the travel demand and land use models.
What is the best methodology to expand the on-board transit survey results to facilitate route level analysis and forecasting?
Expanding transit on-board survey data is a complex procedure that should be done by experienced professionals in conjunction with the Federal Transit Administration.The task could include time-of-day and direction expansion as well.The data could also be cross referenced with income / auto ownership data.
Further manipulation of the data should focus on transit corridors rather than individual routes.On-board transit data can also be cross referenced with household survey data to identify possible new routes or other transit enhancements.
The panel felt that the CHCNGA-TPO should consider collecting on-board survey data for the free downtown shuttle route that was previously excluded from data collection activities.
If transit is incorporated into the mode choice model, it should include a transit network that can have trips assigned to it.This will allow for detailed transit analyses, including by route and stop.
Should the highway re-validation process (only highway component) delay the development of the transit and non-motorized components?
The panel felt that with only a year or so before the model needed to be ready for LRTP modeling that it would be best to focus first on updating the highway component of the model. However, as previously mentioned, all trips should be included in trip generation with the non-motorized trips then being removed.
TDOT is thinking of building a toll bridge under Public Private Partnership: Is it costly that the model reflect this purpose? How difficult is it to incorporate cost into specific analyses (e.g. analyze gas prices affect on transit ridership)?
Before any type of pricing data can be evaluated the travel demand model needs to be updated to include cost factors, such as auto operating costs, parking costs, and value of time.Incorporating these factors is a fairly intense effort; however, the effort required can be reduced by borrowing factors / values from other travel models in the region.These types of model enhancements also serve as a good precursor to implementing a full mode choice model.
If the CHCNGA-TPO considered transitioning to an activity-based model, should parallel tracks of modeling be simultaneously considered? Would it be better to start with an UrbanSim land use model, knowing that it will better integrate with TransCAD and will require developing synthetic populations which will be needed for a future activity-based model?
The panel did not believe that the CHCNGA-TPO should consider transitioning to an activity-based model in the short term, although it may be applicable in the long term.Instead near and medium term model development efforts should focus on enhancing the existing four-step model. When the CHCNGA-TPO does begin to move towards an activity-based model, a hybrid approach such as that used by the Knoxville Regional TPO may be applicable.
Should assessments of the reasonableness of transit and non-motorized traffic modeling be by route, stop, and time of day for transit?How would similar assessments be performed for the non-motorized component?
The panel felt that transit reasonableness should be evaluated at the route level, but not necessarily at the stop level.They also didn't see a need for time of day transit modeling, but did suggest considering peak / non-peak modeling.
For non-motorized travel modeling, the panel recommended that an off-model sketch planning tool be used on a sub-area / corridor basis.
Some Transportation Demand Management strategies are supposed to be part of the evaluation process; How can they be evaluated?
These types of strategies are difficult to model with a regional travel model and are probably best left to off-model sketch planning tools.
How can the decisions of the various stakeholders that influence land use development be modeled without relying on a micro-simulation discrete choice model?
In the short term, consider forming one or more focus groups consisting of business and development interests, economic development groups, chief planning officials, utility service providers, etc.Use these focus groups to develop a list of growth drivers, their priority, and relative importance among each other, which can be used in the land suitability analysis of CommunityViz.
Considering the current data available, staffing, and timetables, what are the major barriers to development of a detailed (micro-simulation) regional land use model?
As mentioned in the panels' general comments, it is felt that a micro-simulation land use model (e.g. UrbanSim) should not be considered in the short- term, but should be the subject of further research for possible long-term implementation.This recommendation is based on the effort that is necessary to get a properly functioning simulation-based land use model up and running. Some MPOs have spent years on this process.Given the need to begin producing forecasts for the next LRTP in 2012 it is not feasible to implement UrbanSim before then.
How can the land use model be optimized to reduce run time while also maintaining the quality of the modeling?
The panel felt that run time could be a metric used by the CHCNGA-TPO in comparing the type of land use and / or modules to be utilized. If the CHCNGA-TPO proceeds with using CommunityViz, they could consider varying the unit of analysis (i.e. parcel of grid cell) or size of units in some modules in of the model to increase efficiency.For example, large recreation or otherwise development-constrained homogeneous areas could be represented as a single large cell rather than multiple small cells, which would reduce the internal model calculations necessary and speed run time.
Integration of Land Use and Transportation Models
How can the travel demand model and land use model best be integrated when they are running on two different software systems?
The panel did not feel that software differences would be barrier to model development and application.It is a fairly simple exercise to export a DBF, CSV, or other file from one model and import it into another.
What accessibility measures should flow from the TDM to the LUM?How can intrazonal travel, transit, pedestrian, and bicycle accessibility be measured and fed into the LUM?
In the short-term, performance measures and/or assumptions embedded in the CommunityViz land use model will provide some answers, but without a mode choice component to the travel demand model, the best accessibility measure for the suitability analysis is probably the roadway volume-to-capacity ratio.Over time as the LUM and TDM models become more complex, measures such as the presence or absence of a travel mode, congested travel time, cost, and logsums could also be utilized.
What demographic measures can flow from the LUM to the TDM?Are there other measures from the LUM that could be utilized as inputs to the TDM?
The demographic measures that can be expected to move from the LUM to the TDM are the standard socioeconomic data inputs:households, population, household size, employment by category, etc.As cost factors are introduced to the TDM, income will also need to become an input.The key is to match the level of detail coming out of the LUM to the data requirements of the TDM.As mentioned previously, there is little value in building great complexity and detail into the LUM only to aggregate the results to simplified values for use in the TDM.
Other possibilities with the CommunityViz framework are tying TDM inputs to the form and pattern table for each place type.
How can the integration between the LUM & TDM be validated / tested? How does the CHCNGA-TPO know the integrated model is "working properly"?
The key, as with any model calibration / validation exercise, is to ensure that the models accurately represent base year conditions."Stress testing" can be utilized to test sensitivity to certain changes, which results can be compared to expected values to help judge the reasonableness of the model results.Another factor to be considered is the adequacy of the convergence of the feedback loop between models.