Title: Distribution of epicenters in the Olami-Feder-Christensen model

Abstract: We show that the well established Olami-Feder-Christensen (OFC) model for the dynamics of earthquakes is able to reproduce a new striking property of real earthquake data. Recently, it has been pointed out by Abe and Suzuki that the epicenters of earthquakes could be connected in order to generate a graph, with properties of a scale-free network of the Barabasi-Albert type. However, only the non conservative version of the Olami-Feder-Christensen model is able to reproduce this behavior. The conservative version, instead, behaves like a random graph. Besides indicating the robustness of the model to describe earthquake dynamics, those findings reinforce that conservative and non conservative versions of the OFC model are qualitatively different. Also, we propose a completely new dynamical mechanism that, even without an explicit rule of preferential attachment, generates a free scale network. The preferential attachment is in this case a ``by-product'' of the long term correlations associated with the self-organized critical state. The detailed study of the properties of this network can reveal new aspects of the dynamics of the OFC model, contributing to the understanding of self-organized criticality in non conserving models.