Historically, in cellular networks, the user equipment (UE) association with a base station (BS) has been largely confined to nearest BS association strategies. Unlike cellular networks wherein the BSs are static, vehicular networks are prone to high mobility, resulting in the UE experiencing a Doppler shift. The Doppler effect further results in the fading channel exhibiting high temporal variability. In this work, we present a novel stochastic geometry framework by modeling the effects of the Doppler shift on the physical layer network performance through the coherence time. First, the statistics of the Doppler shift are studied in a two- and three-dimensional Poisson point process (PPP) setting. The distribution and the geometric factors influencing the Doppler are analyzed. Next, we propose Doppler aware utilities, modeling the Doppler induced Shannon rate achievable in time-selective fading links. Through these utilities, we study the outage and coverage probabilities for a UE to achieve a given quality of service. These probabilities are computed based on the optimal BS association in vehicular networks, which go beyond the closest Euclidean distance. The novel BS association factors in the UE-BS line of sight, the BS direction of motion, the fading gain, in addition to the Euclidean distance. We show that the nearest BS is no longer the optimal BS association, resulting in non-convex coverage tessellations associated with the BSs. Our simulation results illustrate the need for a Doppler correction, complementing the Shannon performance. The network performance achieved through the proposed Doppler-Shannon based association is proved, analytically as well as through simulations, to be superior to the classical Shannon based association achieving gains up to 12% and 55% in a 2D and 3D setting, respectively.