In this paper we will show a self-organizing hierarchical topology that can serve as infrastructure for efficient and reliable safety related communication aiming to minimize the interference between network participants. We will show how reliability, fairness, and efficiency can be achieved in our presented D-CUT algorithm. Our solution addresses these challenges in a local, distributed manner by exploiting the vehicle proximity map, needed from a safety point of view, as the building block for constructing the hierarchical topology. The D-CUT algorithm produces a geographically optimized clustering of the network, by grouping dense and consecutive nodes into clusters which are separated by maximally possible gaps. This type of clustering allows strong connections between cluster members and reduces the inter cluster interference. In addition, we present the primitives for interference aware communication system design, based on the awareness of vehicles to their surrounding vehicle proximity map partitioned into geographically optimized clusters. We present theoretically provable bounds demonstrating the ability of the algorithm to deal with dynamic nature of the VANET environment supported by simulation results.