The University of Utah
College of Architecture + Planning

Key Enhancements to Conventional Four-Step Travel Demand Models

There are four major contributions that the team hopes to develop that will significantly improve the travel demand modeling (TDM) process.  The first of these is a calibration and validation of the vehicle ownership component of the model.  Previous methodology utilized a multinomial logistical regression, which the team believes to be inappropriate in this application.  Instead, the researchers will be employing a Poisson or negative binomial regression to estimate vehicle ownership.  Secondly, a calibration and validation of the intrazonal travel model will be applied.  Comparisons of household travel survey data to TDM outputs show a significant under-prediction of intrazonal travel by the model.  Intrazonal travel is a trip in which the destination and origin are both within the same traffic analysis zone (TAZ).  Third, the team will conduct a calibration and validation of the walk and bike mode choice model.  Previous research has used limited data in the study of this aspect of the TDM.  Using a database of household travel survey data from fifteen regions throughout the country, it is expected that a much more robust model will be generated.  Also, the current model combines walk and bike trips into a single unit; these modes will be separated.  Finally, a peak spreading model will be incorporated.  Peak spreading is a phenomenon in which increased congestion during the peak commuting hours influences individuals to choose to travel at different times to avoid the delay, thus shifting the demand of roadways to new periods.

This project will build upon the existing TDM used by the Wasatch Front Regional Council, utilizing a database created by the researchers at the University of Utah.  This database is particularly noteworthy in that it is the most ambitious endeavor to date to combine household travel survey data from numerous regions throughout the United States.  The database includes 850,000 trips from 91,000 households.  This high volume of data will add to the validity as well as the generalizability of the findings.