Assessing Freeway Expansion Projects with Regional Dynamic Traffic Assignment
Corresponding Author: Norman Marshall, Smart Mobility Inc.
Presented By: Norman Marshall, Smart Mobility Inc
Benefits of freeway expansion often are overstated in alternatives analyses. There are dozens of examples across the U.S. where major expansion projects have failed to reduce travel times, or have done so for only five years or less. Post hoc explanations include inaccurate land use forecasts, and/or failure to account for induced travel. However, there often is a more fundamental explanation – a poor modeling protocol.
Project assessments generally rely on regional travel demand models using Static Traffic Assignment (STA). These models often forecast future traffic flows exceeding capacity upstream and downstream of the study area. For example, the California SR 710 North Extension Project full tunnel alternative model shows an impossibly-high average traffic volume of 2214 vehicles per lane per hour for the 13-hour period from 6 AM to 7 PM upstream of the project. Metrics from such a regional model exaggerate expansion benefits in the study area, and underestimate expansion impacts outside the study area.
Replacing STA with Dynamic Traffic Assignment (DTA) in the regional model produces more realistic metrics for freeway expansion projects. In a test of the I-30 freeway expansion project in the Little Rock region, the regional travel demand model with STA and feedback estimated that the expansion project will reduce Vehicle Hours Traveled per day by 3,869 in 2040. In contrast the regional model with DTA estimated that the project will increase VHT per day by 2,451 in 2040 because VMT was higher, and there was little change in regional average speed.
The common practice of exporting link volumes or subarea trip tables to microsimulation does not resolve this problem. In the Little Rock example, an environmental study exported inflated one-hour 2040 trip tables, and estimated that the project would reduce VHT by 12,152 in the AM and PM peak hours alone. The inflated trip tables created unrealistic traffic backups in the No build alternative. Despite the added complexity and cost of microsimulation, the two-model approach resulted in the least accurate forecasts of the three methods. The most accurate forecasts were achieved using the regional model with DTA.