Abstract

This paper describes a new spatio-temporal access method (SEST-Index) that combines two approaches for modeling spatio-temporal information: snapshots and events. This method makes it possible to not only process time slice and interval queries, but also queries about events. The SEST-Index implementation uses an R-tree structure for storing snapshots and a log data structure for storing events that occur between consecutive snapshots. Experimental results that compare SEST-Index and HR-tree show that, for a change frequency between 1% and 13%, SEST-Index requires less storage space than HR-tree, and for a change frequency between 1% and 7%, SEST-Index outperforms HR-tree for interval queries. In addition, as SEST-Index is an event-oriented structure, event queries are efficiently answered. In order to decrease the storage space for frequencies of change above 20%, this work explores alternatives that optimize the space of the log structure without affecting the efficiency of query answers. Copyright 2005 ACM.