Many Ubiquitous computing applications can be considered as planning and acting problems in environments characterised by uncertainty and partial observability. Such systems rely on sensor data for information about their environment and use stochastic or probabilistic reasoning algorithms to infer system state from sensor data. We propose a domain modelling technique that characterises the sensor and actuator infrastructure and the set of system states in Ubicomp application domains. We capture the location and geometry of all domain model elements and use these spatial properties to tailor system state determination to reflect the quality and spread of the sensor and actuator platform. The domain model is used to support the development of Ubicomp applications and allows us calculate the degree of observability that exists over the state space in the environment. The degree of observability over the state space can act as an input into determining the state inference and action sele...