Thurston Regional Planning Council (TRPC) Travel Model Peer Review Report
7.0 Panel Discussion and Observations
This section summarizes the panel's answers to TRPC questions and offers the peers' overall observations. TRPC staff already have a healthy appreciation of the uses and limitations of models, a strong consultant team ready to get started, and budget for the upcoming household survey. The agency has in the past demonstrated a sensible approach to model development and application. This discussion can thus proceed quickly to a focus upon short- and long-term suggestions for the model update after a relatively brief treatment of TRPC's analysis needs. While this section covers much of the same material found in the previous section's point-by-point response to TRPC questions, it pivots that material into a potential action plan.
7.1 Observations: Analysis Needs
7.1.1 General Comments
TRPC has a precise understanding of its modeling needs and priorities including its responsibilities under federal and state law plus its locally-driven analysis requirements (see Section 4.2). Its current model appears to have amply served the legally-mandated applications. The panel expects that judiciously-implemented updates will allow the model to continue doing so in the future. TRPC's stated analysis priorities make sense both in terms of the mandated applications and the locally-driven applications.
7.1.2 Panel Observations on TRPC Analysis Priorities
The high priority TRPC places on updates, enabling better modeling of interregional travel flows and their impacts on I-5, is sensible. Anecdotal remarks from TRPC and partner staff during the peer review indicated that the current model has to be "tricked" into handling future interregional flows along the I-5 corridor by the addition of two I-5 travel lanes (one each direction) which do not now exist and for which there are no future plans. Such tricks benefit neither general planning/air quality applications nor investment study modeling. Since the most direct solution (expanding the model boundary appropriately) has a profound effect on the model and supporting data gathering, it makes sense for TRPC to treat this as a high-priority analysis need.
Another main TRPC analysis priority-mobility alternatives analysis on I-5, with managed-lane/pricing alternatives-is related to the general priority of addressing interregional travel. Keeping this as a priority likewise makes sense for two reasons: successfully modeling I-5 is part of the solution to the general interregional issue; and enabling meaningful analysis of pricing and managed lanes will drive other fundamental model update tasks in useful directions.
TRPC's third stated priority follows logically from the above. The potential for transit and carpool/vanpool solutions to help with the interregional and I-5 issues is well worth assessing given environmental and budgetary pressures on potential action alternatives in the I-5 corridor.
The main comment the panel made about TRPC analysis priorities regards what the agency listed as "Corridor Analysis": studying possible operational solutions on existing and potential "Smart Corridors"; and studying related system/demand management options, in general. This corridor analysis drives the stated intent of enhancing TRPC's DTA capability, which is sensible. The panel observed in addition, though, that as important as it is to do what one can to treat managed lanes properly in the demand model, to really understand them operationally and when priced, the demand model is insufficient. DTA has demonstrated its ability to address priced managed lane analyses (e.g. US 36 in Colorado) and TRPC already has both existing DTA capabilities plus plans to enhance those capabilities. The panel therefore suggests using DTA for managed lane analysis, where appropriate, to supplement the demand model.
7.2 Potential Shorter-Term Model Updates
The panel observed that TRPC could focus on the following steps in the near term:
- Expanding the model boundary to "internalize", in all model components, all Puget Sound Region geography that produces or attracts significant amounts of travel to and from the TRPC region. This should be in close coordination with supporting data acquisition tasks, especially the design and scope of the upcoming 2012 TRPC household travel survey and its supplementation using data from the 2006 PSRC survey. Specifically, this would entail:
- Choosing the new boundary based on careful understanding of existing travel flows (the 2010 I-5/US101 and the 2006 PSRC HH surveys being prime sources of such information) so that all significant movements are internalized.
- Carefully building in consistent demand and network treatments of "special generators" such as JBLM and government campuses using observed performance data (gate count, traffic count), asset data (lanes, capacities, etc.), and (for JBLM) a targeted sample from the 2012 TRPC HH travel survey.
- The potentially useful (but optional) additional enhancement of replacing passenger work trip generation/destination choice with a usual workplace location choice model. This could potentially be estimated from 2010 Census data, the 2010 OD Survey, and appropriate American Community Survey (ACS) Census Transportation Planning Package (CTPP) data. PSRC has an operational submodel of this type that could be used for comparison.
- Establishing active coordination with PSRC on all model expansion tasks, but especially special generator treatments.
- Carefully coordinating the design and implementation of the new 2012 TRPC model survey by:
- Extracting data on trips to/from the TRPC model area from the existing PSRC HH survey.
- Targeting samples of the JBLM population and the travelers currently considered to be "outbound commuters."
- Borrowing the PSRC survey design for the survey to be conducted in the TRPC region (but also being careful to ensure that all necessary details are captured for all respondents, such as trip end location, no matter where the trip starts and stops).
- Carefully designing targeted sampling and deploying multi-modal instruments (phone, mail, web) in the survey to ensure complete sampling of representative demographic groups.
- Seeking to have a transit on-board survey done by Intercity Transit in a time frame compatible with the TRPC 2012 HH survey.
- Improving the model's sensitivity to pricing analysis with a series of individual enhancements. Note that these enhancements are also likely to help with the boundary expansion task since high-priority model applications are intended to analyze pricing in the I-5 corridor which carries much of the travel that TRPC desires to "internalize."
- Establish a more-detailed and consistent treatment of value-of-time (VoT) in all model components. TRPC has two basic alternative approaches to this:
- Option 1 (the "basic" approach):
- Borrow PSRC's 11 assignment classes (HBW x 4 income categories, etc.) or something similar.
- Carry the income classes through the entire model chain
- Be careful in application since recent research has found that low income travelers can have very high VoT in some cases, meaning that aggregate approaches like this entail inaccuracies.
- Option 2 (the "state of practice" approach):
- Again borrow PSRC's 11 assignment classes or something similar.
- Replace the current trip generation submodel with a population synthesizer (see Joan Walker's paper on microsimulating a 4-step model)
- Use distributed VoT from the population synthesis in mode choice and assignment; even if the distribution is asserted, it will be more useful than the static alternative.
- Ensure accurate representation of peak-hour LOS in all modeled time periods. Without this, the model is less sensitive than it should be to pricing and congestion.
- Build LOS sensitivity into trip generation to handle latent demand and flexible work response to pricing and congestion.
- Feed LOS and accessibility into the auto ownership submodel. For example, think of the concept of location-efficient mortgages.
- Add a time-of-day choice model in conjunction with moving to peak period assignments: this makes the model sensitive to demand management policies and pricing strategies in ways that static proportions from base year diurnal factors cannot. The time-of-day submodel should be responsive to the peak-hour LOS (some agencies like PSRC use smaller time periods such as half hours for their time-of-day submodel), and is a useful way of addressing peak spreading issues. Without time-of-day choice, the model will not be responsive to variable pricing.
- Strongly consider applying the TRPC DTA capability for understanding fine-grained time-dependent response to congestion, where appropriate. The demand model will not have the time granularity of the DTA.
- Steps A and B will naturally benefit from a solid foundation in the form of proper representation of appropriate network detail. Within the context of the demand model, it is helpful to properly represent capacity, especially in pricing analysis. Two-way-left-turn lanes (TWLT), auxiliary lanes, and hard-shoulder running are instances of real-world solutions that are sometimes coarsely represented in demand model networks. Given that TRPC intends to update its network representation as a part of this overall effort, the panel made some specific suggestions regarding these special features:
- Arterial geometries with TWLT and other features can be represented by developing from observed traffic data facility-specific volume-delay functions (VDFs). A facility typology might be something like 5-lane arterial, 4-lane arterial, 3-lane arterial, 2-lane local, etc. Alternatively, coding TWLTL as a half-lane would be consistent with PSRC's current practice.
- Hard shoulder running could be treated by the use of an additional fraction of a lane in the modeled time periods when the shoulder is open to traffic; the additional fraction would depend upon the facility type, shoulder geometry, permitted speed, or similar factors.
- Auxiliary lanes can be treated as a full additional lane if the network is coded so that every change in overall cross section starts another model link. Some agencies use a fraction of a lane to represent the auxiliary lane if they choose not to create separate links every time the facility cross-section changes.
7.3 Potential Mid-Term Model Updates
The shorter term possibilities listed above are mutually supportive and in several cases (e.g. time-of-day choice and peak-hour LOS representation) highly interdependent. In a certain sense, they make a complete set and they also address what the panel understands to be TRPC's highest-priority analytic needs. The mid-term potential tasks are those that would be useful to TRPC's analytic priorities but that are not quite as wrapped up together with the potential short-term items listed above. The following steps are more self-contained (and therefore easier to do later):
- Enhance the model's treatment of carpool/vanpool and general transit modes.
- A key enhancement is a more representative treatment of zone-to-zone transit LOS in destination choice and mode choice. This should be achieved by incorporating transit, walk, and bike in the utility logsums.
- It would be very useful to fully treat Park and Ride (PNR) and drive-access to fixed-route transit in the model. This can help the desired enhancements to the modeling of carpool/vanpool use of PNR facilities. INRO supplies pre-packaged PNR macros that can help. In addition to internal trips, be sure to represent PNR access to northbound and southbound trips now treated as "external-internal" or "internal-external" in--collaboration with PSRC--to provide accurate LOS skims.
- Evaluate the potential of treating "park and pool" travelers (the vanpoolers/carpoolers using PNRs as rendezvous parking locations) as fixed-route transit, since such arrangements are typically scheduled in advance and repetitive. The region's Commute Trip Reduction (CTR) data can help here.
- Do proceed to differentiate walk and bike access to transit by:
- Building in sufficient network representation to accurately represent walk access in the new, denser TAZs.
- Treating walk-, bike-, and auto-access separately with appropriate "travel sheds" and speeds.
- Upgrade the TRPC DTA model to encompass the existing and anticipated subareas to be analyzed for operational, managed lane, or system management purposes.
- Start by coding all base year freeways with all lane details; this provides the common reference framework for all further subareas and enables managed-lane analysis in the DTA model.
- Next code all base year network details across all analysis subareas.
- With the required network in place conduct a basic (not fully detailed) calibration on the entire system. This will provide insight into any major demand/supply inconsistencies (since the DTA model is capacity-constrained), and if required possible implications for the demand model can be considered. Coding options for intersections that are outside of the detailed study corridors can also be considered at this point, e.g. as was done in the Smart Corridors project.
- Note that, although coding effort is low relative to calibration effort, it is crucial to have comprehensive high quality network data for coding AND high quality observed performance data for calibration. Good data on both topics reduces the calibration effort required.
- As individual studies commence, perform detailed calibration and validation in the study area. Note that adding minor network detail should not require major recalibration.
- The panel also recommended crafting institutional arrangements so that TRPC is notified of major operational changes that would require larger-scale updates and more detailed re-calibration in the future (see San Diego's arrangement for distributed system data entry: the TRB Planning Applications conference in Reno in 2011 had a presentation on it by Joaquin Ortega). Overall network updates can then be data-driven on an as-needed basis.
- The panel suggested approaching sponsor agencies for the individual studies not only for data but also for model development and calibration funding.
7.4 Potential Long-Term Model Updates
The panel observed that it may be sensible for TRPC to implement freight model updates later, in the longer term. The justifications for this observation are indications from TRPC materials and verbal responses to panel questions that (a) much Port of Olympia traffic is specialized (log-export related), (b) there are some but not a high number of freight distribution centers in the region, and (c) a good proportion of freight movements in the region are "through" movements. Regions with more sophisticated freight models by comparison have (a) larger ports (or similar special generators) serving diverse commodities, (b) many distribution and transfer centers, and (c) large amounts of internal freight productions/attractions in addition to through movements. Freight is still important to model; it just might not be as important as other features at this time. When TRPC does invest in its freight capabilities the panel recommends the relatively simple approach described below.
- Enhance the model's freight treatment using existing data and national parameters (i.e. Quick Response Freight Manual, second edition).
- Focus improvements on major TRPC planning needs by ensuring that freight flows on I-5 and US 101 are accurate. To this end, consider engaging WSDOT for data collection and even model development and calibration support given the statewide importance of these facilities.
- Use more detailed employment categories in truck trip generation than simple total employment. The Production, Distribution, and Repair (PDR) category should capture distribution centers, but it is important to verify that the employment data used for truck trip generation is accurately categorized.
- On a separate topic, TRPC should consider partnering with PSRC, WSDOT, and other western Washington agencies to conduct a "Cascadia" travel survey covering both PSRC and TRPC regions together. One way of accomplishing this would be to do an "add-on" to the next PSRC HH survey and for PSRC to support an "add-on" to the proposed 2012 TRPC survey.