Pre-aggregation in Spatial Data Warehouses

Data warehouses are becoming increasingly popular in the spatial domain, where they are used to analyze large amounts of spatial information for decision-making purposes. The data warehouse must provide very fast response times if popular analysis tools such as On-Line Analytical Processing [2](OLAP) are to be applied successfully. In order for the data analysis to have an adequate performance, pre-aggregation, i.e., pre-computation of partial query answers, is used to speed up query processing. Normally, the data structures in the data warehouse have to be very “well-behaved” in order for pre-aggregation to be feasible. However, this is not the case in many spatial applications.In this paper, we analyze the properties of topological relationships between 2D spatial objects with respect to pre-aggregation and show why traditional pre-aggregation techniques do not work in this setting. We then use this knowledge to significantly extend previous work on pre-aggregation for irregular data structures to also cover special spatial issues such as partially overlapping areas.