Ohio-Kentucky-Indiana (OKI) Regional Council of Governments Peer Review Report
6.0 Peer Review Panel Recommendations
The following section describes the panelists' recommendations pertaining to OKI's three major topics for review.
Review of the Current Trip-Based Model
The following suggestions were made based on the review and discussion of OKI's current trip-based model's structure, procedures, and forecasts:
- The roadway network file should be shared with ODOT for review and commentary, particularly for the identification of coding errors.
- Market segmentation in trip generation should be focused on significant travel behavior differences and less reliant on workers and auto-ownership. The panelists recommended that income be used for segmentation.
- Trip distribution model testing should be conducted through simple travel time impedance tests rather than log-sums.
- Trip distribution needs to be calibrated based on the chosen segmentation, i.e. income, and carried through mode choice to help with the tolling element of the model.
- An auto-ownership model is recommended for mode choice and auto sufficiency.
- Mode choice coefficients and nesting structure should be revisited, and different nesting structures should be considered. FTA best practice documentation was identified as a primary resource for this transition, including NCHRP 716, Appendix A (http://onlinepubs.trb.org/onlinepubs/nchrp/nchrp_rpt_716.pdf) and a presentation on Travel Forecasting for New Starts presented at a workshop in St. Louis, Missouri in September of 2007: (http://www.fta.dot.gov/documents/Sessions_01-04.pdf).The panelists noted that nesting coefficients should not be set to one (1.0) unless a multinomial logit model structure is desired.
- New mode choice coefficients should be reviewed by plotting the transit assignment results of the sensitivity tests for a light-rail scenario, a scenario that OKI has analyzed in the recent past and with which they are familiar.
- Survey findings related to peaking effects and traffic counts by time-of-day should be compared to the temporal factors developed for the model to confirm that the time period lengths accurately represent regional peaking phenomena.
- Trip assignment and validation should be revisited and adjusted in accordance with documentation of best practices, including NCHRP Reports 255, 365 and 716, as well as the ODOT Certified Traffic Manual.
- Additional validation for DynaSim should be conducted, specifically by saving the paths of the model. It was suggested that traffic counts also be used.
Data Improvements and Collection Efforts
The following list provides the panelists' reactions to the data sources currently employed by the OKI model, as well as the identification of additional, readily-available data sources for consideration in the existing model and for planned improvements.
- A consistent transit boarding and alighting data resource should be identified and implemented.
- The 2010 transit on-board sample requires summary and refinement. Before the survey is incorporated into the model, trip table assignment should be checked and compared to FTA path finding standards.
- Additional analysis of the 2010 OKI household survey results:
- The household survey should be compared to data from the survey conducted in 1995 at a comparable level, i.e. - trips or tours, and any significant variation should be reviewed and explained. The survey should also be compared with other recently conducted surveys in the state and in peer areas.
- Results should be summarized in production-attraction format and recalculated based on trip length distributions. It was suggested that, at a minimum, home-to-work distances be tabulated.
- The processes used in the household survey data mining work should be revisited to potentially enhance trip-purpose categorizations. For example, rather than applying block group average vehicles per household, distribution could be performed by distributing results into bins of zero, one, two, and three-or-more vehicles per household.
- BEA and LEHD sources should be used to distribute employment by NAICS code to provide a comparison to the current employment data in the model. It was noted that the OKI model relied heavily on detailed employment categories for attraction rates; therefore, this data check is critical. BEA data will also be helpful to provide consistency in data across the three states. When employment is allocated to TAZs, it should be expanded to match the BEA data totals. Additionally, InfoUSA data should be purchased for year 2015, as well as each subsequent year. The InfoUSA data can be compared with Quarterly Census of Employment and Wages (QCEW) data and LEHD data.
OKI Response to Employment Data-Related Panelist Suggestions:
OKI's employment data for seven of it's eight counties is developed from actual business records and supplemented as necessary with information from other sources so all employers are included. For the eigth county (Dearborn) OKI purchases InfoGroup data which consists of individual business records. OKI geocodes all of these records and aggregates employment to each TAZ. OKI has found that many businesses in the raw data have been associated with the wrong county, particularly where the business location is near a county border.OKI catches and corrects these errors in the geocoding process.
OKI does review and compare county-level employment data from the BEA, as well as the Bureau of Labor Statistics (BLS). OKI feels that its data is more accurate for trip modeling purposes than either of the other sources. BLS data is built on QCEW data which only includes employment covered by unemployment insurance. BEA data is comprised mainly of QCEW data supplemented with administrative records to include sole proprieters. As stated in the methodology, BEA assumes that place-of-work and place-of-residence are identical for both farm and non-farm proprieters. Therefore, the BEA data is inflated for trip modeling purposes as only commuting employment is pertinent. Both BLS and BEA county-level data contain the inaccuracies of businesses associated with the wrong county.
OKI looked at the LEHD in the past and found it lacking. However, OKI is happy to revisit the source.
OKI will continue to utilize whatever employment data sources are provided by the Ohio and Kentucky departments of transportation. OKI is reliant on InfoGroup data for Dearborn County and will continue to acquire employment data from that source.
- If OKI would like to include HBU as an individual trip type trip type, a survey of local universities should be administered. If this type of survey is not feasible, university trips should be incorporated as an HBO trip type or included in special generators.
- Land use data should be obtained from county tax assessor geodatabases or economic development agencies and subsequently used for household and employment allocation.
- Information from the available statewide models should be used at exchange points for external-external trips and external-internal trips. It may also be used strategically for freight purposes.
- Average growth rates from AirSage should be used at all external stations. Growth rates should be distributed by directional axes.
- While current freight estimation practice was considered sufficient, a truck or commercial survey in the region was recommended to improve estimation.
Guidance Regarding Transition to Tour-Based Model
The panelists' primary guidance with regard to the transition to the tour-based model was to improve their data resources available, as described in the "Data Improvements and Collection Efforts" bulleted list above. In the long-term, the panelists recommended consideration of region-wide DTA to support further simulation efforts and support tour-based modeling.
In the near-term, OKI will continue to work with the current GPS household data to overcome some of the shortcomings in data processing and address the issues described above regarding the assignment validity.
Specific model improvements are likely to include development of a university-trip model; improvements to trip distribution models, possibly including destination choice models; calibration adjustments in the Dayton region (MVRPC). OKI is also considering updating the regional truck model, adding toll modeling capabilities, dynamic traffic assignment, and integration of non-motorized trip making.
