Study of low-magnitude seismic events near the Novaya Zemlya nuclear test site

A study of available seismic data shows that all but one of the 42 known underground nuclear explosions at Novaya Zemlya have been detected and located by stations in the global seismic network. During the past 30 years, only one seismic event in this area has been unambiguously classified as an earthquake (1 August 1986, mb = 4.3). Several other small events, most of which are thought to be either chemical explosions or aftereffects of nuclear explosions, have also been detected. Since 1990, a network of sensitive regional arrays has been installed in northern Europe in preparation for the global seismic monitoring network under a comprehensive nuclear test ban treaty (CTBT). This regional network has provided a detection capability for Novaya Zemlya that is shown to be close to mb = 2.5. Three low-magnitude events have been detected and located during this period, as discussed in this article: 31 December 1992 (mb = 2.7), 13 June 1995 (mb = 3.5), and 13 January 1996 (mb = 2.4). To classify the source types of these events has proved very difficult. Thus, even for the mb = 3.5 event in 1995, we have been unable to provide a confident classification of the source as either an earthquake or explosion using the available discriminants. A study of mb magnitude in different frequency bands shows, as expected, that the calculation of mb at regional distances needs to take into account source-scaling effects at high frequencies. Thus, when comparing a 4 to 8 or 8 to 16 Hz filter band to a “teleseismic” 2 to 4 Hz band, the smaller events have, relatively speaking, significantly more high-frequency energy (up to 0.5 mb units) than the larger events. This suggests that a P-wave spectral magnitude scale might be appropriate. The problem of accurately locating small events using a sparse array network is addressed using the 13 January 1996 event, which was detected by only two arrays, as an illustrative example. Our analysis demonstrates the importance of using accurately calibrated regional travel-time curves and, at the same time, illustrates how array processing can be used to identify an interfering phase from a local disturbance, thereby avoiding location errors due to erroneous phase readings.