A NATIONAL TRAVEL DEMAND MODEL FOR THE U.S.: A PERSON-BASED MICROSIMULATION APPROACH

Understanding long distance travel behavior and forecasting reliable long distance travel demand are critical in evaluating intercity or regional transportation improvements and infrastructure investment projects. As the nation and various states engage in funding transportation infrastructure improvements to meet future long-distance passenger travel demand, it is imperative to develop effective and practical modeling methods for long-distance passenger travel analysis. This dissertation proposes the first integrated activity-based travel demand model system for individual’s quarterly/yearly long distance or national activities and travel in the U.S at the Metropolitan Statistical Area (MSA)/Non-MSA level. The model system is developed based on a rigorous behavioral framework in long distance travel planning, and takes into account the specific attributes of the long distance travel such as low frequency, long activity duration, different sets of mode alternatives, etc. The system includes three tiers: 1) the yearly long distance activity pattern level estimating the number of different activities a person will choose during one year; 2) the tour level which consists of tour destination choice, time of year choice, tour duration, and tour mode choice; 3) the stop level estimating the intermediate stop frequency, purpose and location. According to the different decision-making processes for different types of long distance activities (business, personal business, and pleasure), two tour-level model structures were developed, one for long distance business/personal business activities and the other for long distance pleasure activity. Econometric model developments are conducted for the multiple model components. And estimation results are obtained based on the 1995 American Travel Survey data, transportation origin-destination (OD) skim data, and economic/demographic data. With-out-sample validation is performed for each model component and system-wide model calibration is conducted using optimization method prior to model implementation and future year policy analysis. The model system is implemented in our developed micro-simulation platform which simulates each individual’s yearly long distance activities and travel in the U.S with the input of the population data, the associated transportation OD skim data and economic/demographic data. The travel demand in the year of 2040 is forecasted and two more scenarios including national-level fuel price increase and high speed rail operation are analyzed based on the calibrated long distance travel demand model. The contributions of the dissertation lie in the following three aspects: 1). The first national travel demand model which employs a person-based microsimulation approach is developed for the U.S. for long distance passenger travel analysis; 2). the developed person-based travel demand model enables us to conduct the travel demand analysis of high speed rail in selected inter-regional corridors in the U.S and the national-level fuel price increase; 3). a post-processing learning system which can estimate the missing information such as trip purpose for the passively collected long distance travel survey data is proposed and tested