The relationship between urban environments, residents, and transportation system is not static, but dynamic within a day, from one day to another, across weeks, seasons, and years. However, many transportation models are developed based on the assumption of a "typical day", which creates issues of credibility in both forecasting travel demand and assessing transport policies. In this dissertation, the different levels of travel behavior dynamics are examined using a variety of sources, and the three levels of dynamics are defined based upon time scales as follows: Microdynamics -- 24 hours e.g. daily activity travel patterns, Macrodynamics -- life long longitudinal travel behavior, Mesodynamics --

travel behavior with some regularity on a weekly/monthly/yearly basis. I address the impacts of life events, interpersonal interaction, and spatial characteristics on travel behavior at spatio-temporal scales ranging from neighborhoods to states and from within a day to decades. In order to examine these dynamics, I use different statistical models including spatial econometrics, structural equation models, and latent class analysis. First, I show the significant role of human interactions in explaining time-of-day dynamics of activity travel behavior and its relationship with life-cycle stages, socio-demographic characteristics and dynamic patterns of activity participation and experienced accessibility. In terms of longitudinal dynamics, I examine 13 years of dynamics of activity travel patterns and changes in household composition, employment status, and land use are found to be significant triggers of behavioral changes.

In addition, I also developed econometric and spatial models to help utilize social media data and passively collected GPS data in travel behavior analysis. Using Twitter data to examine its usefulness in travel behavior analysis, I found both the advantages and disadvantages of using these data for transportation research by comparing them with statewide trip records from travel survey data and traditional transportation models' outputs. Twitter data are useful in estimating activity space and its growth patterns over time when heavy users' tweets are used for smaller area. The findings reveal dynamic links among them that are neglected in static models and can be incorporated in a new generation of behavioral models.