This work presents an evaluation of vehicular mobility traces acquired from GPS device in the Muriaé city, state of Minas Gerais, in order to characterize the movement of buses in the public transport system. Statistical data were extracted from the following components: velocity, acceleration, angle of direction change and pause time. From the obtained data, we found out Weibull distributions which showed a good fit to both velocity and angle of direction change components. Moreover, Normal and Lognormal distributions were found to represent, respectively, acceleration and pause time components. Thereafter, we used the vehicular mobility traces to carry out a performance evaluation of routing protocols proposed for delay tolerant vehicular networks, then it was observed that the MaxProp showed better performance in the scenarios evaluated. Finally, the information obtained on the real mobility was used to configure and adapt some known vehicular mobility models. Then, these models were used to investigate through simulations, the impact of the real mobility and the synthetic mobility of the network performance. It was concluded that the synthetic mobility not satisfactorily replace the real mobility when the intention is to make an accurate analyze of the performance of routing protocols for vehicular networks.