Our vision of a viable way for transparent and meaningful processing of heterogeneous spatio-temporal data is to put data semantics in the foundation of an integration process. We present and correlate means of integration as components of the mediation level of an interoperable system. For our domain of interest we present MADS domain ontologies and MADS conceptual data model dedicated to modeling of spatio-temporal data. Using as example two MADSschemas we outline an integration methodology based on semantic interschema correspondence assertions and integration goals.