ADOT requested that the panel provide insight and guidance related to topics of interest for the AZTDM3 and AZTDM4 development. These topics helped frame the panel discussion and recommendations. The topics provided by ADOT in this peer review are as follows:
The following text summarizes a point-by-point response by the peer review panel to the topics of interest posed by ADOT.
The panel recommended rechecking calibration targets on individual model steps when using updated skims to ascertain if the model is still within validation targets. The panel commented that recalibration should be performed if changes to the impedance matrices prevent individual model steps from meeting the validation thresholds for that step. The panel was not aware of a set of thresholds or criteria governing changes in impedance tables that dictate when to redo a model calibration. The panel also recommended using averaging in feedback loops, if it is not currently being employed. Averaging trip tables or impedances helps to dampen oscillations and/or drifts that may occur in successive iterations of the feedback loop.
The panel recommended that ADOT adjust the Census Transportation Planning Products (CTPP) journey-to-work data that was used to calibrate the work-trip purpose in the long-distance model to be in the same units as home-based work trips. CTTP reports on work trips based on the respondents' usual trip patterns and does not account for absenteeism and working from multiple locations. Home-based work trips represent an actual trip made (or not made) on the survey date. The panel commented that the adjustment between these two definitions may be significant.
The panel recognized that many of the limitations to the long-distance model stem from a lack of data. The panel recommended that ADOT first seek to better understand long-distance trips by obtaining more data in order to improve the long-distance model. To do this, the panel offered the following suggestions:
The panel commented that rural areas near urban areas may have different travel patterns because of access to the urban areas. As a result, ADOT may want to consider clustering samples in these cities, or other smaller cities, in order to obtain a sufficient sample size to adequately understand trip behavior. The panel also commented that the NHTS may not conduct long-distance surveys; however, it may be possible to recruit a household during the NHTS for an additional survey if that household took a long-distance trip in the recent past.
The panel commented that ADOT would need to determine the importance of non-resident trips. Since non-resident trips are typically not identified in a household travel survey, assumptions may need to be made about non-resident trips if data cannot be found.
The panel noted that ADOT could apply correct methods to the long-distance model once long-distance trip behavior is better understood. Specifically, the panel recommended considering separating long-distance, home-based work trips from short-distance work trips in the model; this would better account for the long, inter-city commute trips. The panel also suggested that ADOT consider additional terms in the destination choice model to capture intercity interactions, such as between Tucson and Phoenix. The panel also suggested that additional market segmentation in trip distribution, where there is a closer match between employment and job types, may help.
The panel commented that many agencies, such as California DOT, have recent experience in collecting data from long-distance surveys. These agencies could provide feedback to ADOT regarding their experience and also advise on potential pitfalls. The panel recommended that ADOT contact agencies that recently completed trip-diary or long-distance surveys to better understand the types of questions they asked and the lessons they learned from their data collection efforts.
The panel also recommended that ADOT investigate the use of other sources to quantify the magnitude of long-distance trips (both resident and non-resident), including through cell phone data, which could then be controlled using other data sources, such as CTPP data. The panel commented that while cell phone data would help to quantify the magnitude of travel, ADOT would probably still need survey data to understand the details about the trips since cell phone data is aggregate in nature. The panel commented that smartphone survey applications with GPS tracking may be the future of trip surveys, but acknowledged these are still in beta testing.
The panel also recommended that ADOT consider vehicle body type identification and location tracking technology at weigh-in-motion (WIM) stations to help identify long-distance truck movements for the freight model.
The panel recommended that ADOT consider an AB model for the long-distance component first and leave the short-distance models as trip-based models. The panel commented that aggregate-demand models typically work well in evaluating highway projects, which affect all users in the same way. Aggregate-demand models are less effective for projects that affect user groups differently, such as managed lanes, where tolling could affect people in different ways or at different times of the day.
The panel also commented that the traffic modernization evaluations ADOT wishes to perform could be achieved through the use of DTA-whether AB models or trip-based models were being used. Enhancements to the demand models may not be necessary to address operational traffic issues.
In reference to DTA, the panel recommended consideration of zonal-level data in addition to the network detail that ADOT is planning to collect in the AZTDM4. Large zone sizes typically found in statewide models and more moderate zone sizes typically found in MPO models will likely lack sufficient resolution needed to capture the effects of intersection improvements. A much finer spatial resolution and supporting data would be needed. The panel commented that the multi-resolution approach ADOT is proposing for sub-area modeling may be a good way of getting more refined data for traffic operations analyses. The panel recommended that ADOT consider prioritizing updates to the long-distance and freight models before DTA.
The panel acknowledged ADOT's efforts to improve aspects of the truck model in the AZTDM3. The panel noted that the primary uses of a statewide model are to estimate long-distance and freight trips and that plans for further development of the freight model in the AZTDM model improvement plan were missing (other than an update to incorporate FAF county-to-county data). The panel recommended that ADOT consider further freight model enhancements in their model development plan.
The panel also recommended that ADOT consider adding an auto sufficiency parameter in addition to an auto availability parameter in the model. Auto sufficiency is defined as the ratio of potential drivers to the number of vehicles available in a household. The auto sufficiency variable might also be measured as the number of vehicles versus the number of workers. This measure could be a better parameter to use in the mode choice model.
