Comparative Seismic Detection and Location Capabilities of IRIS and GSETT-3 Stations Within Central Asia

Danny Harvey, University of Colorado, Boulder

Our ability to confidently detect and accurately locate seismic events will form the backbone of the International Monitoring System, known as the IMS. This system will provide the crucial verification element of the new comprehensive test ban treaty. Stations of the Global Seismographic Network will be an important component of the IMS. In addition to providing an operational IMS component, IRIS, through its Joint Seismic Program, has conducted network and array deployments in Central Asia. These deployments have been used to evaluate a proposed IMS primary array site and, as described here, to evaluate overall IMS detection and location capabilities within a region of the world that will be important for the CTBT.

The final technical test of the IMS, known as GSETT-3, started on January 1, 1995 and has been operating continuously since that time. The final seismic product of GSETT-3 is known as the Reviewed Event Bulletin (REB) which is intended to be a comprehensive and accurate listing of worldwide seismic activity produced within 48 hours of real time. This bulletin would be used by the international community to detect potential treaty violators and it would set into motion the treaty enforcement procedures, starting with on-sight inspection. An accurate determination of worldwide REB detection thresholds and true location errors is critical for assessing overall IMS monitoring capabilities and effectiveness.

Although many network simulation studies have been done to estimate IMS detection and location capabilities, the most reliable and straightforward method for making these assessments is to compare the REB with bulletins determined from other networks. When local or regional networks are used for the comparison, we can get a good fix on the local seismicity characteristics and use this to pin down the REB capabilities within the region.

Even in cases where the reference network is global in scale, a direct comparison of the REB with the reference bulletin can be useful. This can be seen in Figure 1 which shows worldwide seismicity for 11 months of 1995 using the GSETT-3 REB and the USGS Preliminary Determination of Epicenters (PDE) bulletin. We compared the two bulletins and only plot events that were seen in one bulletin but not the other, with red indicating events that were in the REB but missing from the PDE and green visa versa. We can see that in comparison the PDE has a lower detection threshold than the REB in certain areas of the world, such as Southern Europe and Western US, and the REB has a lower detection threshold than the PDE throughout most of the world generally.

Figure 1. A comparison of REB and PDE bulletins for 11 months of 1995. Only events that appear in one bulletin but not the other are shown.
The IRIS JSP has operated several networks in Central Asia over the last five years. We have used the data recorded by these networks, along with GSN and CDSN data within and around Asia, to compile a regional bulletin which we call the Central Asian Bulletin (CAB). Because of the relatively high density of stations within Central Asia, the CAB can be used as a reference bulletin for determining the REB detection and location characteristics for the region. Figure 2 shows a direct comparison of REB and CAB bulletins for the month of February, 1995. Green symbols represent events that are in the CAB and not in the REB and red visa versa, with the common events not plotted. It is apparent that the CAB has a lower detection threshold than the REB over a large region of Central Asia.

Figure 2. A comparison of CAB and REB bulletins for February, 1995. Only events that appear in one bulletin but not the other are shown.

A more quantitative determination of Mb detection threshold values can be seen in Figure 3. This figure shows the numbers of events as a function of Mb magnitude for all shallow events (depth 50 km) within a 10 degree radius of the Kyrgyz station at Ala-Archa. The fall off in numbers with decreasing magnitude indicates the detection threshold value. For the REB this detection threshold is about 4.3 and for the CAB it is about 3.5.

Figure 3. Numbers of events as a function of Mb for the CAB, REB, and PDE bulletins. Only crustal depth events within a 10 degree radius are shown.
Detection threshold characteristics are important performance parameters for a network. Equally important for nuclear monitoring are the location error characteristics. A large error in estimating source depth can put a legitimate suspect event out of consideration. Large errors or uncertainties in the geographic location can make on-sight inspection difficult or impossible. In Figure 4 we show a comparison of event locations for February, 1995 between the CAB and the REB. The location error ellipses are plotted with the REB ellipses not filled and the CAB ellipses filled with green. A line is drawn between the CAB and REB hypocenters for each event.

Figure 4. A comparison of location error ellipses for events from the CAB and REB bulletins. The green filled ellipses are those from the CAB bulletin and the unfilled ellipses are those from the REB bulletin.

We note that the location error ellipses are much larger for the REB than for the CAB. In some cases the REB error ellipses are so large that making an on-sight inspection would be practically impossible. Also we note that in many cases the error ellipses for the REB and CAB locations do not intersect, indicating that one or both location and/or error estimates are inaccurate. In particular, the event that lies closest to the Kyrgyz Network, the cluster of triangles (stations) in northern Kyrgyzstan, shows a small CAB error ellipse and a much larger REB error ellipse that does not contain the CAB error ellipse. Because of the proximity to the Kyrgyz Network, in this case we feel certain that the CAB location and error estimate are more accurate and that, consequently, the REB location and/or error estimate is inaccurate.

A stated goal for monitoring of seismic events is that the location error footprint be less than 1000 square km in area. In Figure 5 we show the location error area as a function of Mb magnitude for events within the REB and CAB. The symbols in green are the CAB events, which represent events in February, 1995, and the events in red are the REB events in the same month. Only shallow events (depth 50 km) that are within 10 degrees of the station at Ala-Archa, Kyrgyzstan are plotted. Since there were relatively few REB events during the month, all of the REB events for the year of 1995 are also plotted as light red squares. The triangles show mean log statistics with the dark green and dark red vertical bars showing the standard deviations.

Figure 5. A comparison of location error area for events from the CAB and REB bulletins. Red colors represent REB events and green colors represent CAB events. Mean and standard deviation statistics of the log of the area are shown by the triangles and vertical bars.

Although many individual CAB events exceed the 1000 square km threshold, the CAB mean statistics stay within the threshold throughout the magnitude range. For the REB, both individual events and the mean statistics exceed the threshold for events below magnitude 5. A curious feature of Figure 5 is the final upturn of the CAB mean error for the largest magnitude events. This comes about because of the increased numbers of stations that contribute to the locations for large events. These stations become distributed over a worldwide region and as they start to see the larger events they actually degrade the location solution by introducing a large number of distant observations that effectively reduces the importance of observations from the closest stations. In order to compute accurate error footprints in the magnitude 3 to 5 range, source location computations need to account for changes in the receiver constellation that will occur in this magnitude range. We have shown how data collected from IRIS deployments can be used to assess the performance of the prototype IMS. These data can also be used to predict how enhancements to the IMS would effect performance. As the CTBT comes into force and the IMS moves into an operational mode, IRIS facilities will continue to provide the raw material for both operational and evaluational functions in support of the test ban.

Return to: IRIS Newsletter Information
Return to: Title Page and Table of Contents
Return to: Previous Article

Continue to: Next Article