A Relational Algebra Approach to Problems of Spatio-temporal Representation

This paper develops an approach for the representation of natural world phenomena within a relational structure. Relational algebra (RA) as the basis for database has been enormously influential and successful in the utilisation of computers in the world of business applications. This success has led to attempts to utilise relational algebra for application of computers to problems of the natural world with varying degrees of success, relational algebra underlies the database of most GIS. In current applications it does not deal well with many features of the natural world. These issues include the representation of features such as plains within valleys, relative representations, the representation of context knowledge, multiple instantiations of elements. The difficulties of addressing these issues within the RA database approach have lead many authors to suggest alternative architectures and relational algebra has become an unfashionable tool in this area. We take a contrary view and instead propose a method that uses the features of the natural world to build relational data structures from both prior and discovered knowledge of the natural world. This is developed from a consideration of the application of relational algebra to the business model. In this case the database issues discussed are rare and fall under the ambit of nulls. In the business case the model is constructed such that nulls (and the like) are considered as exceptions and dealt with on an individual basis. The key point in that the structure is sufficient as to render this problem a minor one. We propose a system that makes use of the structure of the database to include nulls and multiple representations to develop a structure that will support representation of the natural world within a relational structure. This paper takes the form of theoretical discussion with real illustrations and some practical testing.