Multivariate analysis of microscopic traffic variables using copulas in staggered car-following conditions

The current study investigates the applicability of the copula approach to capture the potential dependence among microscopic traffic variables (time headway, speed and centreline separation) and the joint probabilistic behaviour of the variables in staggered car-following conditions, using trajectory data from video recording of a four-lane divided road in Kolkata, India. As importantly, the study indicates that although Frank, Clayton and Gaussian copulas provide better fits for the two-dimensional traffic variables; considering both the positive and negative associations among the three microscopic variables, and checked by the goodness-of-fit measures based on Rosenblatt's transformation, the Gaussian copula is found acceptable to model the multivariate joint distribution and is hence, employed to construct the joint and conditional probabilities among the three variables. Overall, the findings of the study justify the need to explicitly consider centreline separation and vehicle speed in time headway modelling for elucidating the integrated driving behaviour of non-lane-based traffic streams, a proper consideration of which will further ameliorate the realistic replication of rider's behaviour in simulation modelling.