Accurate assessment of earthquake risks requires an accurate inventory of the property at risk. Typically, the inventory data must be assembled from widely scattered secondary sources such as assessor's parcel files, U.S. Bureau of the Census population and housing information, and aerial photographs. The aggregation of the data into a usable form is tedious, time- consuming, and difficult. Furthermore, the use of these heterogeneous data sources invariably results in missing data and in some cases inventories which are of unknown reliability. As computerized techniques for evaluating seismic hazard continue to evolve, we will need to develop methods for assembling integrated, accurate inventory databases. This project will develop methodologies and computer tools for the automatic integration of inventory databases for seismic risk assessment. A knowledge-based system will integrate multiple, differing databases into a single unified data structure for representing seismic inventory data. By incorporating automatic integration, the system will provide more up-to-date, more complete data for use in seismic risk assessment. Furthermore, new computer programs for risk assessment will need to deal only with the unified data structure rather than with the myriad of local data formats, thus increasing the programs' applicability. In conjunction with the knowledge-based integration, we will develop an expert system to address the problems of incomplete and inconsistent data within individual and multiple databases. The expert system will formalize the heuristics and rules-of-thumb that seismic experts use in evaluating and reconciling inventory data.
