Article Preview
TopIntroduction
Data warehousing and On-Line Analytical Processing (OLAP) systems are technologies designed to support business intelligence. OLAP models (hypercubes) are based on the concepts of dimensions and facts (Inmon, 1996). A fact is a concept that is relevant for the decision-making process, and it is described by a set of numerical indicators (measures). Dimensions, composed of hierarchies, allow for the analysis of facts along different analysis axes at different levels of detail.
New information and communication technologies make it possible to collect huge amounts of geographic data. These data are generated by remote sensing systems or other computer applications (Franklin, 1992). Geographic information is described by two components (Longley et al., 2001). The spatial component is the geometry and its position on the Earth’s surface. The semantic component is a set of (1) descriptive attributes and (2) spatial, thematic and map generalization relationships. Geographic Information Systems (GIS) have been developed in order to store, organize, visualize and analyze geographic data, (Longley et al., 2001).
Spatial analysis aims to understand, estimate and predict real phenomena, showing recurrent spatial structures and shapes. Several spatial operators have been proposed (e.g., overlay, map join, etc.) but a “standard” model and allied algebra have not yet been defined (Voisard & David, 2002). Nevertheless, Longley et al. (2001) proposed a classification of spatial analysis operators, i.e., query and reasoning methods, measuring methods, transformation methods, and synthesis methods. Transformation methods modify geographic data (i.e., overlay, buffer, etc.) through logic and/or spatial rules. Query and reasoning methods exploit relationships between geographic objects to enable multigranular spatial analysis (Timpf & Frank, 1997; Camossi et al., 2008). Here, data are represented at different levels of detail (or ‘granularity’), i.e., cities and regions, etc., to allow for support spatial analysis by adding or downscaling details for particular datasets through zoom-out/zoom-in operations.
Therefore, a new kind of Decision Support Systems called Spatial OLAP (SOLAP) has been developed in order to effectively factor spatial data into multidimensional analysis. SOLAP tools integrate GIS functionalities (memorizing, analyzing and visualizing) into OLAP and data warehousing systems (Marketos et al., 2008; Rivest et al., 2005). SOLAP tools were recently successfully used to analyze agricultural, economic, seismological data (Marketos et al., 2008), etc.
These systems are based on spatio-multidimensional models composed of spatial dimensions, spatial measures, which are analyzed through spatio-multidimensional operators. Multidimensional models have been proposed for SOLAP (Ahmed & Miquel, 2005; Pourrubas, 2003; Jensen et al., 2004; Damiani & Spaccapietra, 2006; Gómez et al., 2009; Sampaio et al., 2006; Silva et al., 2008; Glorio & Trujillo, 2008) which formalize the concepts of spatial dimensions, spatial measures and spatio-multidimensional operators. In particular, they define spatial measures as geometric values and/or the result of spatial operators. Spatio-multidimensional operators are defined as extensions of OLAP operators for spatial dimensions.
However, in our opinion, the existing SOLAP models actually limit certain aspects of spatio-multidimensional capabilities, namely the semantic component of geographic information and flexibility of spatial analysis. When geographic information is used as measure, SOLAP models reduce it to geometry without taking into account its relationships. Consequently, they support multigranular analysis, but with some limitations.