In dynamic networks, topology changes require frequent updates of centrality measures. Such perpetual calculations range from computationally hungry to unfeasible before the next topological change happens. On top of this, the centrality may seem unstable or even random from the nodal perspective, making them difficult to predict. In this paper, we propose to shift the focus of centrality estimation from the conventional topological perspective to a geographical perspective. We advocate that some centrality metrics are, depending on the situation, inherently related to the geographic locations of the nodes. Our strategy associates a measure of centrality to coordinates and, consequently, to the nodes that occupy those positions. In a vehicular scenario, the one we consider in this paper, geographical centrality is much more stable than node centrality. Hence, nodes do not have to compute their centralities continuously – it is enough to refer to their geographic coordinates and find the centrality by merely consulting a pre-established table. We evaluate our strategy over two large-scale vehicular datasets and show that, whenever we match a centrality to an area, we correctly estimate the centralities up to 80% of the highest-valued nodes.