Data warehouses keep large amounts of historical data in order to help users at different management levels to make more effective decisions. Conventional data warehouses are designed based on a multidimensional view of data. They are usually represented as star or snowflake schemas that contain relational tables called fact and dimension tables. A fact table expresses the focus of analysis (e.g., analysis of sales) and contains numeric data called measures (e.g., quantity). Measures can be analyzed according to different analysis criteria or dimensions (e.g., by product). Dimensions include attributes that can form hierarchies (e.g., product-category). Data in a data warehouse can be dynamically manipulated using on-line analysis processing (OLAP) systems. In particular, these systems allow automatic measure aggregations while traversing hierarchies. For example, the roll-up operation transforms detailed measures into aggregated data (e.g., daily into monthly sales) while the drill-down operation does the contrary. Data warehouses typically include a location dimension, e.g., store or client address. This dimension is usually represented in an alphanumeric format. However, the advantages of using spatial data in the analysis process are well known since visualizing data in space allows users to reveal patterns that are difficult to discover otherwise. Spatial databases have been used for several decades for storing and managing spatial data. This kind of data typically represents geographical objects, i.e., objects located on the Earth’s surface (such as mountains, cities) or geographic phenomena (such as temperature, altitude). Due to technological advances, the amount of available spatial data is growing considerably, e.g., satellite images, and location data from remote sensing systems, such as Global Positioning Systems (GPS). Spatial databases are typically used for daily business manipulations, e.g., to find a specific place from the current position given by a GPS. However, spatial databases are not well suited for supporting the decision-making process (Bédard, Rivest, & Proulx, 2007), e.g., to find the best location for a new store. Therefore, the field of spatial data warehouses emerged as a response to the necessity of analyzing high volumes of spatial data. Since applications including spatial data are usually complex, they should be modeled at a conceptual level taking into account users’ requirements and leaving out complex implementation details. The advantages of using conceptual models for database design are well known. In conventional data warehouses, a multidimensional model is commonly used for expressing users’ requirements and for facilitating the subsequent implementation; however, in spatial data warehouses this model is seldom used. Further, existing conceptual models for spatial databases are not adequate for multidimensional modeling since they do not include the concepts of dimensions, hierarchies, and measures.
Only a few conceptual models for spatial data warehouse applications have been proposed in the literature (Jensen, Klygis, Pedersen, & Timko, 2004; Timko & Pedersen, 2004; Pestana, Mira da Silva, & Bédard, 2005; Ahmed & Miquel, 2005; Bimonte, Tchounikine, & Miquel, 2005). Some of these models include the concepts presented in Malinowski and Zimányi (2004) and Malinowski and Zimányi (2005), to which we will refer in the next section; other models extend non-spatial multidimensional models with different aspects, such as imprecision (Jensen et al., 2004), location-based data (Timko & Pedersen, 2004), or continuous phenomena such as temperature or elevation (Ahmed & Miquel, 2005).
Other authors consider spatial dimensions and spatial measures (Stefanovic, Han, & Koperski, 2000; Rivest, Bédard, & Marchand, 2001; Fidalgo, Times, Silva, & Souza, 2004); however, their models are mainly based on the star and snowflake representations and have some restrictions, as we will see in the next section.