The following text summarizes the current version of the Arizona Statewide Travel Demand Model (AZTDM2) at the time of the review, along with data sources used in the development of the model.
In AZTDM2, the travel market is divided by person trips and truck trips. Each is also sub-grouped by distance. For person travel, 50 miles is the threshold of short trip versus long trip. Truck travel is sub-grouped based on the judgment if the truck trip crosses FAF3 zone boundaries.
The AZTDM2 model structure includes separate components for passenger travel and truck trips as shown in the following flow chart:
AZTDM2 has five primary model components as described in the current model documentation:
Setup/Trip Generation
In this stage, trip generation is conducted for short distance person trip only. Truck and long distance person trips are processed separately in other stages of the model.
Trip generation rates for short distance person trip were generated based on the 2009 National Household Travel Survey (NHTS). For each county, rates were calculated for five trip purposes:
Person trip generation rates are stratified by the area type. Area type definitions were calculated based on an accessibility measure. Area types used by AZTDM2 are:
Skimming
AZTDM2 creates toll/non-toll skims using a generalized cost that is based on travel time, toll and distance for the following four vehicle classes:
Person Travel
In this stage, AZTDM2 performs trip distribution for short distance person travel using a destination choice logit model. After distribution, transit trips are factored off by purpose and area type. A multinomial logit model is then used to predict auto occupancy. Shares were derived from NHTS and then smoothed to ensure logical relationship among modes.
Truck & Long Distance Person Travel
In this stage, the model separately processes short & long distance truck travel and long distance person travel. Short distance truck model is a three-step model without mode choice. Its trip generation is segmented by twelve land use categories:
A gravity model is applied to distribute short distance truck trips. Friction factors between zone pairs are calculated dynamically based on congested travel time.
Long distance person trips are mainly processed by a Java script, which reads and expands the 2002 NHTS long distance data to a state-to-state trip table then disaggregated to TAZ using household data, employment and a weighting scheme. A 10% sample of ticked air travelers by BTS was also used. After missing NHTS records are synthesized and the NHTS data are expanded, trips are disaggregated to the AZTDM zones based on population and employment. The model also uses state parks as a special attraction.
Long distance truck trips are also processed by a Java program but uses FAF3 data to create FAF-district to FAF-district commodity flow matrix, which are then disaggregated to TAZ based on employment. Long distance commodity flows are converted to truck trips using payload factors for single-unit and multi-unit trucks. An empty truck rate is used to factor the truck trips for returning empty trucks. Capacity and volume/delay function curve parameters were obtained from MAG. Passenger car equivalent vales were obtained from the Highway Capacity Manual.
Assignment
Highway assignment in AZTDM2 combines the long and short passenger and truck trip tables and assigns them onto the network by four time periods:
The model performs a feedback loop from trip generation to assignment. In first feedback loop iteration, long distance trips (person and truck) are loaded onto network using an all-or-nothing assignment. Then the short distance persons and trucks are loaded to the network with a user equilibrium traffic assignment. Congested travel times are calculated using a BPR-type volume-delay function. Only the AM and MD travel times are fed back. Convergence is reached when the percent RMSE for the AM and the MD periods are both less than 1%. On the final model iteration, the model assigns PM and NT trips to highway network.
Model calibration focused on the state highway system. Traffic counts were used to validate AZTDM2.
Aggregate volume to count comparisons showed that the R-Square for the total flow was approximately 0.97, and the percent RMSE of total flow was 30%.
Screenline validations were also performed. The percent error of volume on 9 of 10 screenlines was within 16%. Variances on the one screenline outlier in northeast AZ were associated with tribal reservations and parks. State line crossing screenlines had the total volume percent error within15% at each border crossing.
The following are the various sources for model development, calibration and validation used in the AZTDM2 model (source: ADOT Peer Review Presentation):
NOTICE
This document is disseminated under the sponsorship of the U.S. Department of Transportation in the interest of information exchange.The United State Government assumes no liability for its contents or use thereof.
The United States Government does not endorse manufacturers or products. Trade names appear in the document only because they are essential to the content of the report.
The opinions expressed in this report belong to the authors and do not constitute an endorsement or recommendation by FHWA.
This report is being distributed through the Travel Model Improvement Program (TMIP).
