Request for TMIP Peer Review
Jim Sullivan, Research Analyst
University of Vermont Transportation Research Center email@example.com
1. Purpose of TMIP Review
The Vermont Agency of Transportation maintains a statewide travel demand model which has recently been updated to a 2009 -2010 base year. Vermont is one of about 2/3 of the states in the US with a statewide travel model. Being one of the smallest states in terms of population and home to only one metropolitan planning organization, Vermont's travel model is a critical tool for travel estimation and forecasting statewide and in its non-MPO regions. The Agency of Transportation has demonstrated its commitment to travel modeling at the statewide level by funding an add-on to the 2009 National Household Travel Survey (NHTS) so that its model could be effectively updated to its current base year.
The University of Vermont Transportation Research Center (TRC), under contract with the Division of Policy, Planning, and Intermodal Development at the Vermont Agency of Transportation (VTrans) , has hosted, improved, and applied the Vermont Travel Model ("the Model") since 2008. The purpose of the Model is to estimate travel demand and link flow throughout the state and across its borders using Vermont's generalized spatial demographics and employment. The model estimates daily passenger-vehicle flow between 866 internal and 70 external traffic analysis zones (TAZs) for four trip types. Commercial truck traffic is estimated directly from traffic counts as a fifth trip type in the Model.
The Model employs a traditional four-step process, using rates and coefficients derived primarily from the 2009 NHTS. Previous NHTS sampling in Vermont had not been robust enough to use for an effective Model update, as compared with sampling from larger states. With this in mind, the TRC and VTrans purchased an add-on to the 2009 NHTS dataset, which resulted in approximately 1,700 Vermont households and over 3,800 individuals surveyed.
The resulting per capita sample representation was among the best in the US for the 2009 NHTS. The resulting tabulation of travel behaviors from this dataset forms the basis of the sub-modules used in the Model. Residential information from both the 2006-2010 American Community Survey (ACS) and the 2010 US Census is used to input household distributions and characteristics in the state. Data from the 2009 Bureau of Economic Analysis (BEA) and the 2009 Vermont Department of Labor QCEW were used to disaggregate employment characteristics among the internal TAZs.
The TRC recently completed the Model update to its current 2009-2010 base year, and VTrans is now considering a host of potential improvements moving forward to better suit its expected uses of the Model. Given the lack of any previous peer review and the advancements made in the travel modeling field in the last 10 years, TRC and VTrans staff feel that a peer review is timely. It is critical at this time for the state to move forward with a model that suits their needs and makes the best use of the funds available for improving the model to accommodate future planning requests like energy efficiency evaluations or air- quality assessments.
2. The State
866 internal and 70 external TAZs form the geographic basis of the Base -Year 2009-2010 Vermont STM. Vermont in 2010 contained approximately 621,000 residents and 256,000 households. The state contains 20 distinct Census urban areas but Vermont's 14 counties are predominantly rural, with the exception of Chittenden County, which accounts for approximately 24% of households, 30% of statewide employment 39% of the internal TAZs, and its largest urban area, Burlington. Chittenden County is also home to the state's only metropolitan planning organization, which uses its own, more finely-resolved travel-demand model. Total employment figures for the state (403,311) are categorized into five user-specific industry groups: Retail (46,116), Manufacturing (32,813), Non-manufacturing (241,912), Education (30,787) and Government (51,683). The highway network maintained by the state consists of 5,250 miles of roadway, and in 2010 an estimated 7.4 billion annual vehicle- miles of travel (VMT).
3. History of the Model
Efforts to develop the first travel model for the state of Vermont began in the 1990's with processes run in the SAS Model Manager 2000 platform and the road network maintained in the TRANPLAN software format. The base-year 2000 Model was improved in 2007 by transitioning to a GIS -based model framework using the CUBE software package. Further enhancements were made to improve the correlation between model outputs and validation data. In the fall of 2008, the TRC began to host the Model, where further enhancements have included :
In addition to these enhancements, the TRC updated the Model to a 2009 - 2010 base year by updating:
The Model uses a traditional four-step process to estimate link flows from TAZ-based population and employment. The trip generation sub-module combines these TAZ-based characteristics with the town-based fractions of cross-classified number of household members and workers to calculate home-based trips produced by each internal TAZ. Trip attractions for all purposes and trip-productions for the non-home-based (NHB) purpose are generated for each TAZ using purpose-specific regression equations, each of which utilizes a different set of employment and/or population characteristic. Truck (TRUCK) productions and attractions are taken as a fraction of the NHB trip totals corresponding to the fraction of trucks in traffic counts in the TAZ.. Total non-TRUCK external person trips (PTs) are then subdivided by the other four trip types using an external trip-fractions table estimated from the NHTS data. The trip production- output is held constant while attractions are adjusted by TAZ to create a balanced trip table.
The passenger trip distribution sub-module takes the balanced productions and attractions table, a matrix of free-flow travel times between TAZs (based on travel at 5 mph over the speed limit, with terminal times of between 1 and 4 minutes for origin and destination) and a set of impedance functions to generate a P-A matrix between all TAZs using the Gravity Model. Since the STM is a daily model, all trips are assumed to return to their origin, necessitating that the final matrix be diagonally symmetric. So the set of trip-specific matrices resulting from the Gravity Model application are transposed, then average with their originals, creating symmetrical matrices with same trip totals as the balanced trip table.
Prior to the network assignment sub-module, the symmetrical person-trip matrices resulting from the trip distribution step are converted to vehicle-trip matrices by applying vehicle-occupancy rates by trip type for internal and external travel (from the NHTS). The assignment sub-module employs a user-equilibrium optimization to distribute trips onto the network, resulting in vehicle flows and congested travel speeds by link for the entire state.
With these improvements, the Model provides the state of Vermont with a useful cost-effective tool to:
4. Assistance in Model Specification and Design
At this time, VTrans seeks a peer review of its Model specifications, design, and plan for future improvement. VTrans trusts that the recommendations of the panel will be innovative, state-of-the-practice ideas and looks forward to including recommended tasks in its work program to enhance Model development and refinement. The panel will also be asked to outline the strengths and weaknesses of the existing Model, and to assist in determining whether application of current or more advanced Models can inform issues specifically faced by the State of Vermont, as deemed necessary by VTrans. VTrans continued commitment to travel modeling is evidenced by its partnership with the TRC in hosting and improving the Model's capabilities. It is anticipated that considerable discussion will take place about the merits of moving towards more advanced sub-modules, and which approach(es) are capable of providing the greatest benefits for VTrans.
How the Models can inform the VTrans's specific desired planning outcomes will be an important topic. The strategic challenges that came out of the Agency's strategic planning work are listed below:
A path and process for continued improvement over the next five years, and priorities for implementation, are particularly sought. Some of the specific improvements that VTrans and the TRC are considering include:
5. Proposed Panel and Availability
Due to VTrans and TRC staff's time constraints with ongoing work, we request the assistance of TMIP staff to put together the members of the peer review panel. The make-up of this panel will hopefully consist of individuals with experience in the issues faced by statewide modelers in the US. TRC staff members have made use of resource materials produced in conjunction with statewide modelers in Connecticut, Florida, Ohio, and Oregon, so it would be ideal if one or more of those DOTs could be represented. In addition, it would be useful to include the following local stakeholders in the Vermont travel modeling community:
The TRC staff contact will be responsible for contacting the local stakeholders. Both TRC staff and VTrans Policy, Planning & Intermodal Development staff will participate in the TMIP review.
6. Schedule and Cost
VTrans would like to conduct the TMIP peer review using video- and telephone-conferencing to avoid the need for travel by panelists. VTrans's responsibilities will be carried out by the TRC staff contact person. Therefore, the only expected commitment from peer review panelists will be time to participate in the review meeting, time to review documents and provide comments, and time to present the recommendations of the review committee.
The TRC staff contact for this peer review has time dedicated to the process in an existing contractual relationship with VTrans. Therefore, we are seeking to implement the process and host a review meeting as soon as possible. Our expectation is that we will be able to begin the implementation of the peer-review recommendations in the next 3-4 months.