Being able to effectively solve spatio-thematic queries is an important pre-requisite for spatially-aware information retrieval. In this paper, we look at the spatial part of such queries, and at methods for an effective reasoning about spatial relevance. We outline a concept for qualitative spatial reference models that are used as the basis for spatial models and spatial reasoning with place names. We use graph-based abstractions of polygonal reference tessellations, together with their intrinsic decomposition hierarchies, to represent spatial reference models in an efficient and highly interoperable way. Reasoning about spatial relevance is based on a metric that evaluates the vertical and horizontal proximity and connectedness of spatial entities.