This paper deals with the problem of 'bringing together' descriptions of approximate spatial location that is specified at different levels of granularity. In this context it will be important to distinguish between the textit{roughness} of the approximation at a given level of granularity and the textit{indeterminacy} that arises when those approximations are transformed to another level of granularity. Indeterminacy arises in particular when a given crisp approximation is transformed (lifted) to a finer level of granularity and when a crisp approximation in a mereologically non-full level of granularity (e.g., a map with 'white spaces' of unknown territories) is transformed (generalized) to a coarser level of granularity. The notion of textit{stratified approximation space} is introduced as the underlying formal framework.