Many solutions proposed in the literature to enforce privacy in presence of location information use, implicitly or explicitly, spatial granularities. However, most of the contributions do not describe the formal and computational properties of this tool in details. In this paper we propose three families of spatial granularities, specifically designed for privacy-aware systems. We mathematically characterize them and prove that all of them have two important formal properties. Then, for each of them, we show how to efficiently compute two fundamental operations.