The first half of the peer review was spent by SRTC staff members making presentations on specific items to the peer review panel. During these presentations, many topics came up which initiated discussion among peer review panel members and between panel members and SRTC. This section documents the key points that arose during these presentations.
The SRTC model is used to answer transportation planning questions and inform a wide variety of policy decisions in the Spokane region such as:
There is an increasing desire by the SRTC board to evaluate policy decisions via a quantitative assessment of the options rather than through qualitative and subjective analysis.
SRTC only identified regionally significant transportation projects for detailed analysis and discussion within Horizon 2040. The model was not used to prioritize projects, and instead many of the projects listed in the current regional plan were carried over from the previous plan. The North Spokane Corridor is currently the only large-scale transportation project under development. It is a limited access highway that will provide a better north-south connection through the region. Recently the last 5 miles of the highway has been funded. There are also a few transit improvements underway such as the planned central city line . Bridging the Valley is an initiative to separate vehicle traffic from freight traffic through the region by removing at-grade crossings. Currently approximately 80 trains/day pass through the region. The Inland Pacific Hub (IPH) project identified transportation related investments to increase economic growth in the State of Washington and Idaho. An economic impact analysis (EIA) and cost-benefit analysis was done on Bridging the Valley as part of the IPH.
The SRTC model is used to evaluate how planned transportation projects will impact the regional transportation system. SRTC reviews the following model outputs of a build scenario compared to a no-build scenario:
SRTC has a desire to develop a toolkit for project prioritization. This toolkit would be used for prioritizing projects in Horizon 2040 as well as application for Surface Transportation Program (STP), Congestion Mitigation and Air Quality Improvement (CMAQ), and Transportation Alternatives Program (TAP) funding. SRTC questioned how the outputs of the model can be better used to evaluate and prioritize projects and what performance metrics can be developed to review the results. The panel suggested to think about how SRTC recognizes a problem and then identify the measures based on what will be useful to SRTC’s policy makers by answering the following questions:
The panel cautioned that metrics developed at the national level may not as useful as metrics developed specifically for SRTC. Many of the national level metrics require a lot of new data collection that may not be feasible for SRTC.
The panel also suggested looking toward other non-model tools to help with project prioritization and long-range planning, as the travel demand model can only go so far. It is important for SRTC to understand the set of tools needed to accomplish the objectives SRTC has for transportation planning. Once the objectives are identified, performance measures can be laid out and then from there SRTC can figure out what can be evaluated via the travel demand model and what should be evaluated using additional tools.
The travel time index is used by SRTC as a metric for identifying the top urban transportation corridors within the region. A travel time index is used to measure congestion. The travel time index is measured as congested travel time over free flow travel time using INRIX data and is used for short-term analysis. The AM peak and PM peak periods are reviewed separately. V/C ratio is not used since the model is not sensitive to it.
Collision data, land-use data, VMT, VHT, and bridge conditions is also used to identify high risk corridors. GIS analysis is also performed using output from the model. The top corridors are identified so that the limited funds can be used to improve these top corridors. However, any capacity increasing project must undergo a strict process to show that other lower cost strategies have been considered.
SRTC also uses the travel demand model to do scenario planning and support sub-area plans. The model is used to support land-use policy analysis, such as evaluating transit-oriented development versus traditional suburban growth. Alternative sociodemographic forecast scenarios are run through the model and travel times and volumes are analyzed. SRTC does revisit the TAZ structure with local jurisdictions on a fairly regular basis to ensure that it matches local comprehensive plans. The panel agreed that the comprehensive plan should be used as a guide for developing land-use forecasts.
Another presentation made by SRTC to the peer review panel focused on assessing issues with the current model. The discussion focused on the household travel survey, trip generation, traffic assignment, and transit assignment.
The most recent regional household travel survey was conducted in spring 2005 with a sample size of more than 1,200 households. Since 2005 the socio-demographics and the travel behavior of the population has changed. For example, families with children have decreased over the past ten years while families without children or non-family multi-person households has increased. In addition, transit ridership with Spokane is experiencing record high ridership. In 2012 when the latest version of the model was calibrated, the household travel survey was reweighted to Census 2010 to account for the changing demographics.
The panelists noted that a household travel survey that has very low regional transit mode shares can pose a challenge when it comes to properly modeling mode choice and transit assignment. The panelists agreed with SRTC that a traditional area of underrepresentation in regional travel surveys are the population groups that use transit, such as the college student population.
SRTC should contact the local colleges and universities to partner in data collection and sharing of travel data. Typically, household travel survey can only reach the college students living in households. The most effective way to reach the other college students living in dormitories and apartments, or with any other arrangements, is to survey them on the attraction end (i.e. the college campus). More specifically, SRTC may collaborate with the colleges to conduct a web-based survey to ask the college students the questions related to their residence type and location, and campus oriented trips. Ten percent response rate should be sufficient to develop fairly reliable statistics, such as residence type and location distribution, trip rates and transportation mode usage. The collected data can be further connected with the other supplemental information, such as on-campus parking, to estimate various models for college student travel. This information should then be used to develop and calibrate a home-to-college trip purpose.
SRTC is concerned that the current base model which was calibrated to Year 2010 conditions based on travel behavior from a 2005 household travel survey does not properly represent the trip generation rates that will occur in forecast years. Similar to the rest of the nation, the percentage of the population that is older will increase significantly from Year 2010 to Year 2040. SRTC was interested to learn if there were other sources of data, besides or in addition to conducting a new household travel survey, that could help with developing new trip generation rates.
The panel agreed that an updated household travel survey is necessary to successfully update the trip generation rates. They noted that the overall trip generation methodology was sound. They also noted that for those socio-economic groups represented in a base year (calibration year) model, it is common practice to not change person trip generation rates from that year to a future year. For example, if the represented group contains all people from a household of size two, regardless of age breakdowns in the household, the same average rate per household that is used in model calibration is also used for any future-year prediction involving households of size two. One approach for exploring the impacts of different assumptions on an average person trip rate is through well-documented scenario planning.
SRTC, during validation, should adjust trip generation rates intelligently to match count data, not through such methods as adjusting the vehicle occupancy rate unless there is survey data to support such a change. Traditional household travel surveys that do not include a GPS component tends to underreport the trips made by people, which will lead to underestimated trip rates. Vehicle occupancy rate, if also derived from the household travel survey, is calculated based on reported trips. Unless it is believed that the non-reported trips have totally different vehicle occupancy rate than the reported ones, vehicle occupancy rate, as a derived statistic from the household travel survey, is relatively reliable. This is why the panel suggests to adjust trip rates, but not vehicle occupancy rates.
For validation, SRTC uses the Travel Model Validation and Reasonableness Checking Manual as a guide. They do collect traffic counts on a regular basis, but are interested in confirming the best location for collecting these counts for ideal validation (i.e. validating to screenlines, cutlines, or cordon lines). SRTC relies on the Washington Department of Transportation (WSDOT) for most of their counts, but in the past have paid to have additional counts collected for model validation. For nineteen out of the twenty-two screenlines the model is under-estimating current traffic levels. The screenlines were collected from a variety of jurisdictions including WSDOT and Idaho DOT.
The panel evaluated in detail the traffic assignment setting within VISUM. They noted that the model had the following issues:
The panel had an extensive discussion concerning whether the conical volume delay function, currently being used in the model, was accounting for delay properly. For most urban arterials, there are two types of delay, link delay and node delay. Link delay is largely determined by the capacity of the roadway and the traffic operating on it (i.e. V/C ratio). Node delay is caused by the traffic controls implemented at the intersection, yield and stop signs and more often traffic signals. Link delay is also the function of link length, but node delay is not. Suppose two identical roadway links of different length, 1 mile vs. 5 miles, carry exactly same amount of traffic and are controlled by exactly the same traffic signals with the same configuration (e.g. cycle length, phasing, G/C ratio), the average travel speed of the shorter link will naturally be lower because its node delay from the traffic signal will take a larger percentage of the total delay. However, BPR and conical functions may suggest the same travel speeds unless the link capacity is a function of link length, which is not very common. A good volume-delay function ideally should have both link delay and node delay components. If, as in many travel demand models, assigning the right amount of traffic on roads is the only role of the volume-delay function, and the accuracy of estimated travel speeds is not a concern, then BPR and conical functions work just fine, particularly for the transportation facilities serving uninterrupted traffic flow, such as freeways and parkways. However, for the model to be useful for project level work, node delays will be necessary in addition to conical delay functions.
The panel was also concerned that if the model is only based on the household travel survey, which does not model certain trips (i.e. freight), then the assignment is missing trips. It may be necessary to intelligently add “fudge factors” to account for missing trips. They suggested a review of the section of NCHRP 365: Travel Estimation Techniques for Urban Planning that discusses the correlation of free flow speed to speed limit. The report suggests increasing the free flow speed by a few miles per hour in relation to the speed limit on highways. While two of the panelists stressed the importance of accurately forecasting travel speeds, one panelist noted that since congestion is not predicted to increase significantly in the future, it is not important to forecast speeds directly. Instead, current condition speeds can be transferred to the forecast years.
With regard to transit validation, the model does a good job at matching system ridership but is not very accurate at the route level. Spokane Transit Authority (STA) currently is testing collecting automated passenger counts (APC), but it has not yet been deployed systemwide. STA uses on-board survey data for planning purposes as well as providing it to SRTC for model validation. STA also has data from their fareboxes that SRTC can use for validation.
SRTC noted that the model is not correctly evaluating the number of park and ride trips, which may be due to the fact that assignment is only conducted for the AM and midday periods, and so return park and ride trips are not easily evaluated. The model currently uses headway, rather than timetable, assignment, and SRTC questioned whether they should move toward timetable assignment as well as all-day assignment.
The panel noted that most travel agencies will move toward schedule-based assignment within the next ten years, but that does not mean that it is the right approach for SRTC. They suggested that SRTC talk to other agencies who are moving toward schedule-based assignment, such as the Delaware Valley Regional Planning Commission (DVRPC), to see if it is an approach that will work for SRTC. The panel stated that with SRTC’s existing software, VISUM, they can try out different settings to obtain a schedule-based approach; however, given the low transit ridership within a region, another tool may be more appropriate for transit project evaluation, such as STOPS. The panel pointed out that the low transit share of 2-3% suggests that these transit riders are mostly captive riders. Therefore, using an advanced transit mode choice model and assignment may not be productive. Instead, the focus of model development should be to build a model that focuses on assigning these captive riders to the correct route.
