- journal_title：Bulletin of the Seismological Society of America
- Contributor：Florence Rivière-Barbier ; Lori T. Grant
- Publisher：Seismological Society of America
- Date：1993-
- Format：text/html
- Language：en
- journal_abbrev：Bulletin of the Seismological Society of America
- issn：0037-1106
- volume：83
- issue：5
- firstpage：1527
- section：Articles
A useful tool for grouping similar seismic events (such as mine blasts) is demonstrated. This tool can be combined with others in a case-based approach to automatically identify and locate the thousands of low-magnitude industrial explosions that are detected yearly by regional seismic stations.
A hierarchical cluster analysis method (complete linkage), using the maximum value of the cross-correlation function as a measure of similarity between event pairs was applied to identify distinct groups of events. The cross-correlation functions were computed between envelopes of the signals recorded on the vertical channel of the central sensor of the FINESA array.
The method was applied to events recorded between 4 November 1990 and 28 June 1991 by the FINESA array and located in an active mining district north of St. Petersburg, Karelia (Russia). One hundred forty-four events were classified into 21 distinct groups. Because ground-truth about event location was lacking in this study, we used mine locations given in the Helsinki bulletin (Uski et al., 1992) and determined from SPOT™ photos to verify the event grouping. For two groups, none of the events was reported in the Helsinki bulletin. For eleven groups of events, at least one event of each group was located at a mine in the Helsinki bulletin; the other events of the groups had either a consistent location, had an automatic location, or were not reported in the bulletin. For eight other groups, the Helsinki bulletin gave two to three different mines as locations for events in the group. Because of this discrepancy, we did not rely on the Helsinki bulletin to associate mines with groups of events. An attempt was also made to associate groups of events with mines using SPOT™ photos. Direct information must be obtained about the exact event location and shooting practices of the mining operations to gain a better understanding of the causes underlying an event classification and to assign each of the defined groups of events to a particular mine.