Analyzing time-varying trip distributions with a random-effect spatial OD dependence model.

This paper proposes a random-effect spatial OD (origin-destination) dependence model to investigate varying trip distributions over time. By proposing a maximum likelihood estimation with spectral decomposition methods, the effects of spatial dependences and the unobservable zonal heterogeneity at the origin and destination can be estimated simultaneously. A series of numerical experiments and a real-world trip distribution study with cellular signaling data collected in Hangzhou, China, are conducted. This paper enriches the existing literature by developing (1) an innovative specification to allow for random effects in existing spatial OD dependence models; (2) an innovative estimation method to obtain the values of parameters and improve model fittings; and (3) a set of numerical experiments and an empirical trip distribution analysis that jointly captures spatial effects (spatial interaction and spatial OD dependences), and the unobservable zonal heterogeneity. This paper can equip policymakers with an effective tool for analyzing the OD travel flow over time which is a groundwork for making appropriate transportation policies.