Abstract Empirical and theoretical analyses of the spatiotemporal dynamics of traffic flow reconstructed from randomly distributed probe vehicle data are presented. For the empirical analysis, probe vehicle data generated by TomTom’s navigation devices in the commercial TomTom’s HD-traffic service as well as road detector data measured at the same road section are used. A stochastic microscopic (car-following) three-phase model is further developed for simulations of a real empirical complex spatiotemporal traffic dynamics measured over a three-lane long road stretch with several different bottlenecks. Physical features and limitations of simulations of real spatiotemporal traffic dynamics are revealed. Phase transition points between free flow (F), synchronized flow (S), and wide moving jam (J) are identified along trajectories of empirical and simulated probe vehicles randomly distributed in traffic flow. As predicted by three-phase theory, the empirical probe vehicle data shows that traffic breakdown is an F → S transition and wide moving jams emerge only in synchronized flow, i.e., due to S → J transitions. Through the use of the simulations, it has been found that already about 2% of probe vehicle data allows us to reconstruct traffic dynamics in space and time with an accuracy that is high enough for most applications like the generation of jam warning messages studied in the article.