Bridging the spatio-temporal semantic gap: A theoretical framework, evaluation and a prototype system

The objective of this research is to formally define a spatio-temporal conceptual model that captures data semantics required for temporal and geospatial applications. We show how the proposed model provides a metaphor that bridges the semantic gap between the real world and its spatio-temporal representation in information systems. Our multi-methodological research approach includes: (i) formally defining a spatio-temporal semantic model called ST USM (Spatio-Temporal Unifying Semantic Model); (ii) evaluating the proposed model using a case study and a laboratory study; and (iii) demonstrating practicality of our proposed model using a proof-of-concept prototype system. We describe a spatio-temporal conceptual modeling approach--applicable to any conventional conceptual model--that incorporates sequenced and nonsequenced space/time semantics. We have applied our annotation-based approach to the Unifying Semantic Model (USM)--a conventional conceptual model--to propose ST USM. ST USM is an upward-compatible, snapshot reducible, annotation-based spatio-temporal conceptual model that can comprehensively capture semantics related to space and time without adding any new spatio-temporal constructs. We provide formal semantics of ST USM via a mapping to conventional USM and constraints (expressed in first-order logic), from which the logical schema can be derived. To evaluate the proposed model, we conducted a case study at the US Geological Survey that helped us assess the extent to which the proposed formalism helps capture all the spatio-temporal data semantics for an application. We show that ST USM is ontologically expressive and leads to schemas that completely capture the requisite spatio-temporal semantics. We conducted a laboratory study and found that an annotation-based approach to capturing the spatio-temporal semantics does not adversely impact the schema comprehension as compared with conventional conceptual models (e.g., USM). This implies that annotations