How Accurate are Travel Forecasts: Back Casting of Truck Lane Restrictions using Multi-Resolution Modeling Methods

The integration of mesoscopic and microscopic simulation models provide expanded dimensions of modeling capabilities by taking the strengths of both model resolutions. Many transportation agencies, practitioners and researchers are beginning to see the advantages of using multiple levels of resolution when analyzing corridor specific problems. Mesoscopic models use dynamic traffic assignment to reroute traffic given various traffic conditions. Microscopic models are used to analyze traffic conditions at the individual car or lane level. Models are calibrated and validated using data collected in the field. Most practitioners validate their models to existing conditions and then forecast future conditions to predict traffic congestion. Once simulation runs are finished, results are presented to hosting transportation agencies and the project is completed. Very few practitioners collect future field data and compare it to the simulated forecasts. This sort of reverse model validation is referred to as back casting. This paper outlines the complete modeling process from model development, conversion, calibration, consistency, validation and ultimately model back casting. A case study involving user class restrictions on Interstate 10 in El Paso, Texas was used to analyze how accurate the models were at predicting future conditions. Researcher’s simulated truck restricted lanes on the freeway to determine how effective the user class restrictions were on overall traffic speeds and travel times. One year later when the truck lane restrictions were in place, field data was collected to determine how accurate the models were at predicting traffic conditions.