Spatial data cubes based on shared dimensions and neighbourhood relationship concepts

Abstract

A data cube has exponential storage and runtime complexities with linearly increasing dimensionality; in addition, a spatial data cube further complicates the issue since it integrates spatial features into a data cube. In this paper, a new data cube approach, named spatial cubing, or simply S-cubing, implements two spatial indexing techniques and two spatial non-relational representations. S-cubing based on shared dimensions is the first non-relational solution designed to support spatial data cubes with continuous dimensions, resolution hierarchies and multiple spatial measures and is capable of running on multi-core computer architectures. S-cubing, based on neighbourhood relationships, implements a new data cube hierarchy algorithm using relationships among cells of a regular grid. Thus, this algorithm creates thematic maps from non-geopolitical regular areas, therein avoiding manual hierarchy definitions. A sequential version is found to be faster than a PostGIS implementation and the parallel version achieved a speedup of 13 with 24 threads.