Location-based applications require a well-formed representation of spatial knowledge. Current location models can be classified into symbolic or geometric models. The former attempts to represent logical entities and their semantics, but requires a large amount of manual effort for describing them. On the other hand, the latter represents the geometric coordinates but not the semantics. In this paper, we present a semantic location model which preserves topology and distance semantics to support location navigation but at the same time facilitates programmatic model construction and maintenance. The model is based on a sound location theory. It is mainly composed of two hierarchies: a location hierarchy and an exit hierarchy, which can be derived from spatial maps, such as floor plans, without manual intervention. Through a series of model construction algorithms and a real example, we show that our model is simple but powerful enough to capture spatial connectivity and hierarchica...