In the Fall 1996 IRIS Newsletter, van der Vink and Wallace (1996) use as one of their examples a small seismic event (magnitude 2.4) on 13 January 1996 near the former USSR test site on Novaya Zemlya. Other seismologists who were requested to evaluate this event obtained a different epicenter location from that of van der Vink and Wallace although they used the same data, namely that from the ARCESS (ARA0) and Spitsbergen (SPA0) arrays. Data from the three-component Norilsk (NRI) station in Siberia indicates that the location determined by van der Vink and Wallace is incorrect and supports the other determinations (Figure 1).
|
| Figure 1: Various epicenter determinations for the 13 January 1996 event at/near Novaya Zemlya: (1) two-array location (SPA0 & ARA0) using IASPEI tables, (2) as for (1) but using Bergen University tables, (3) two arrays and the NRI station using IASPEI tables, and (4) same as (3) but using Bergen University tables. Note, Ringdal´s (1996) two-array location at 75.2°N, 56.7°E is close to our epicenter (4). That of van der Vink and Wallace (1996) would coincide with (1) if we add 5 sec to their Sn-reading. The Novaya Zemlya test sites are marked by stars. The map insert shows the seismic arrays and 3C stations in the vicinity of Novalya Zemlya. Records from APA0 (Apatity) and the new AMD (Amderma) array were not available to us. To our understanding the APA0 array did not detect this event. |
| a) & b) |  | Figure 2: The 3-component NRI (Norilsk, Siberia) records for the 13 Jan 1996 event. The upper and middle parts ) show filtered records in the 1.0 - 2.5 Hz and 2.0 - 5.0 Hz passbands respectively. Notch filters were used in addition to bandpass for removing spectral spikes. In c) the two upper traces are 3C envelopes (Husebye et al 1998) Pn- and Sn- arrival times as picked for this event are marked. |
| c) |  |
The methods for analyzing array-recorded signals are well established and in cases with good SNRs even simple bandpass filtering and subsequent stacking suffice to accurately pick Pn and Sn arrival times. For the 13 January event, this was not the case, due to interference in the SPA0 records with signals from an earthquake on the mid-oceanic ridge west of Spitsbergen (Ringdal, 1996). Van der Vink and Wallace, being unaware of this, picked a too early Sn arrival at SPA0, which resulted in a shorter epicenter distance and hence "moved" the epicenter well into the Barents Sea.
Apart from the misinterpretation of the Sn arrival due to the interfering event, another aspect contributed to the move of the epicenter westward. The differential epicenter angle between SPA0 and ARA0 is only about 45 deg, so the epicenter location resolution is relatively poor in the east/west direction as it depends strongly on the traveltime model used. However, the accuracy improves considerably if relevant Pn and Sn arrival times can be extracted from the NRI records. An additional advantage here is that with three station observations the choice of traveltime table used for location is less critical (Kennett, 1996) if the azimuth coverage is reasonable.
Envelope processing (Husebye et al., 1998) proved successful, as demonstrated in Figure 2c. The top trace is the envelope for the 1.0 - 2.5 Hz passband and a prominent, presumably P-wave arrives at 113 sec. It was also seen in the original records. However, extensive 3C analysis gave an azimuth of about 260 degrees which is far off the 310 degrees expected for Novaya Zemlya and polarization characteristics were non-P. The corresponding waveform has some semblance to a local Rg-wave - no sharp onset and monochromatic pulses. In Figure 2c we also display the 3C envelope for the 2 - 5 Hz passband; a presumed weak P-arrival at 122 sec (SNR ~ 1.3) and a clear, presumed Sn-arrival at 243 sec (SNR ~ 1.5). A SNR of 1.5 is taken to indicate a significant signal in the 3C envelope since this process is similar to incoherent beamforming with a rather long STA window (a threshold of 1.6 is used for NORSAR incoherent beams). As an additional check, we used the Pisarenko et al. (1987) phase-picking algorithm, whose picks coincide with the envelope records (Figure 2c). Using the derived NRI Pn and Sn arrival times, we obtained an epicenter location of 75.38°N, 56.55°E, using the Bergen University traveltime curves (Figure 1). The residuals using the Bergen University traveltime curves are not unreasonably large (less than 1.6 sec), while for the IASPEI model the residuals varied +/- 5.6 sec, although the corresponding epicenter difference was less than 15 km. Note, the weak Pn-arrival in the records are not critical for the epicenter determination; we have also located the event using a Pn arrival at 113 sec (see Figure 2c) and also without the P-arrival of NRI. In the first case the epicenter moved 10 km and in the second case only 2 km (in both cases the local traveltime tables were used).
What can be said about the source of this much debated event of 13 January 1996? Its mb magnitude of only 2.4, which is equivalent to a yield not exceeding 100 tons TNT. Such small charges are typical of many chemical explosions and, even if it were nuclear in origin, it could hardly be termed a nuclear device in a CTBT context, as argued by van der Vink and Wallace. A criterion in favor of an earthquake source is the relatively small P/S-signal ratio (van der Vink and Wallace, 1996). We are not convinced that such a criterion can be used here without extensive modification, because the Barents Sea sedimentary basin blocks Lg-propagation, in other words, Lg-waves leak as S-waves into the upper mantle and reappear in the Sn coda (Mendi et al. 1997). Observationally, this is manifested in an extended Sn-wave train, as seen in Novaya Zemlya recordings.
Using the novel envelope analysis technique we could extract useful traveltime information from the Norilsk 3C station, despite the poor quality of the NRI records for an event of magnitude 2.4 at an epicenter distance of about 1200 km. However, we confess that such low magnitude events can neither be accurately located nor clearly identified with seismological means only. In particular, the IASPEI travel time tables are not well suited for accurate event locations in the Novaya Zemlya area for regional epicenter distances. Clearly, certain event location areas remain politically sensitive, as this event has demonstrated. 
Kennett, B.L.N. (1996): Event location and source characterization, in: Monitoring a Comprehensive Test Ban Treaty, E.S. Husebye and A.M. Dainty (Editors), Kluwer Academic Publishers, Dordrecht, The Netherlands, 501 - 520.
Mendi, C.D., B.O. Ruud, E.S. Husebye (1997): The North Sea Lg-blockage puzzle, Geophys. J. Int., 130, 669 - 680.
Pisarenko, V.F., A.F. Kushnir and I.V. Savin (1987): Statistical adaptive algorithms for estimation of onset moments of seismic phases, Phys. Earth Planet. Int., 47, 4 - 10.
Ringdal, F. (1996): Study of low-magnitude seismic events near the Novaya Zemlya nuclear test site, Semiannual Technical Summary, 1 April - 30 September 1996, NORSAR Sci. Rep. No. 1 1996/97, Kjeller, Norway.
van der Vink, G.E. and T. Wallace (1996): The political sensitivity of earthquake locations, IRIS Newsletter, Vol XV, No 3, 20-23.