Seismic detections of the 15 February 2013 Chelyabinsk meteor from the dense ChinArray
ChinArray is a dense portable broadband seismic network to cover the entire continental China, and the Phase I is deployed along the north-south seismic belt in southwest China. In this study, we analyze seismic data recorded on the ChinArray following the February 15, 2013 Chelyabinsk (Russia) meteor. This was the largest known object entering the Earth’s atmosphere since the 1908 Tunguska meteor. The seismic energy radiated from this event was recorded by seismic stations worldwide including the dense ChinArray that are more than 4000 km away. The weak signal from the meteor event was contaminated by a magnitude 5.8 Tonga earthquake occurred ~20 min earlier. To test the feasibility of detecting the weak seismic signals from the meteor event, we compute vespagram and perform F-K analysis to the surface-wave data. We identify a seismic phase with back azimuth (BAZ) of 329.7° and slowness of 34.73 s/deg, corresponding to the surface wave from the Russian meteor event (BAZ ~325.97°). The surface magnitude (M S) of the meteor event is 3.94 ± 0.18. We also perform similar analysis on the data from the broadband array F-net in Japan, and find the BAZ of the surface waves to be 316.61°. With the different BAZs of ChinArray and F-net, we locate the Russian meteor event at 58.80°N, 58.72°E. The relatively large mislocation (~438 km as compared with 55.15°N, 61.41°E by others) may be a result of the bending propagation path of surface waves, which deviates from the great circle path. Our results suggest that the dense ChinArray and its subarrays could be used to detect weak signals at teleseismic distances.
KeywordsChinArray Russian meteor event F-K analysis
An effective way to detect weak signals in a noisy background is to enhance coherent signals and suppress incoherent noises by stacking the waveforms across a dense seismic array (e.g., Ringdal and Husebye 1982; Rost and Thomas 2002). A seismic array refers to any deployment that has more than three seismometers with the same reference time and instrument response (Rost and Garnero 2004). Using seismic arrays to detect weak events can date back to the Geneva Conference of Experts in 1958 (Mykkeltveit et al. 1990). Selby (2008) analyzed four seismic events recorded at the small-aperture ARCES array in Norway, verifying the detection capability of small-aperture array. More recently, the USArray data have been used extensively for small event detection, such as the North Korean nuclear test on October 2006 (Ammon and Lay 2007), as well as mainshock rupture processes and early aftershocks with the back-projection method (e.g., Meng et al. 2011; Yao et al. 2012; Kiser and Ishii 2013). In this paper, we evaluate the detection capability of ChinArray with array processing techniques (e.g., Vespagram and F-K analysis).
Around 03:20:00 UTC on 15 February 2013, a large meteor entered the Earth’s atmosphere over Russia. The associated bolide exploded subsequently and fragments dropped near Chelyabinsk, Russia (https://en.wikipedia.org/wiki/Chelyabinsk_meteor, last accessed 06/2016). It is the largest recorded meteor event since the 1908 Tunguska event (Ben-Menahem 1975). The equivalent yield of the explosion was about 500 kt of trinitrotoluene (TNT) (Antolik et al. 2014). Many studies focused on the trajectory and speed of the bolide (Borovička et al. 2013; Seleznev et al. 2014). They found that the bolide generated a large shock wave during the last stage of explosion. Others provided detailed observations on long-range infrasound (Le Pichon et al. 2013; de Groot-Hedlin and Hedlin 2014) and surface-wave propagation (Tauzin et al. 2013; Heimann et al. 2013). They suggested that the surface waves were generated by the ground motion coupling from the incident shock waves. Antolik et al. (2014) estimated the location of the Russian meteor event as the source of energy produced by the largest explosion, which is located ~50 km south of the Chelyabinsk. Based on Rayleigh wave observations up to 4000 km away, Tauzin et al. (2013) obtained a surface-wave magnitude of ~3.7. The epicentral distance between the Russian meteor event and ChinArray is more than 4000 km, beyond the distance of previous observations. In addition, the seismic waves from the Russian meteor were interfered by those from a magnitude 5.8 earthquake in Tonga occurred ~20 min earlier (Tauzin et al. 2013). Hence, it provides an interesting challenge for us to test the detection capability of weak signals recorded on ChinArray.
2 Data and array analysis
We note that a large number of waveforms within ChinArray are concentrated between 4500 and 4950 km (Fig. 2b). To achieve a relatively uniform distribution of waveforms across the entire distance ranges, we choose one record with signal-to-noise ratio (SNR) above 16 for every ~20 km, resulting in 51 traces for further analysis. We calculate the SNR for each ChinArray waveform with the signal energy from the Tonga earthquake (the square of the root mean square amplitude across the surface window with length of 100 s) divided by the noise energy (the average energy within the same window length before the predicted P wave arrival of the Tonga earthquake). We choose the Tonga earthquake as references for SNR calculations, mainly because it is relatively difficult to visually identify weak arrivals from the Russian meteor at all stations. To ensure high quality of selected waveforms, we set the SNR threshold to be 16, resulting in 51 waveforms for further analysis. As will be shown later, this selection process helps reduce potential impact of the surface waves from the Tonga earthquake to the surface waves from the Meteor impact.
In order to further confirm the signals from the Russian meteor event, we add 22 seismic stations of Global Seismographic Network (GSN) at distances less than 4000 km (inverted triangles in Fig. 1). Combining the vertical waveforms from GSN and 51 selected waveforms from ChinArray, we identify a clear movement from the Russian meteor event, consistent with a phase velocity v ≈ 3 km/s (Fig. 2a).
2.2 Vespagram and F-K analysis
For the phase in the dark green parallelogram of Fig. 2b, we use the same reference station and the 51 waveforms for F-K analysis. We increase the filter band to 0.05–0.07 Hz for better SNR. Based on the time window from 2770 to 2870 s since Feb 15, 2013, 03:00:00 (03:46:10 to 03:47:50 UTC), we observe the phase with the slowness of 27.95 s/deg and BAZ of 333.4° (Fig. 4b), corresponding to the Russian meteor event (BAZ ~325.97°) based on the USGS location. The BAZ discrepancy is 7.43°, much larger than that of the Tonga earthquake. We find that the beam power in Fig. 4b is not concentrated on a clear point as in Fig. 4a, which is also reflected in a large standard deviation of 142° in BAZ based on bootstrap analysis, indicating that the result is not reliable. This is likely because the Russian meteor event has a smaller magnitude and lower SNR. Hence, the coherence of the Russian meteor phases is not as good as the Tonga earthquake.
2.3 Synthetic test
3 Comparison with the F-net results
4 Location and magnitude estimation
To estimate the surface-wave magnitude M S for the Russian meteor event, we apply the Praha equation M S = log(A/T) +1.66log(Δ) + 3.30 (Karnik et al. 1962), where A is the amplitude of surface-wave displacement in μm after removing the instrument responses, T is the period of the corresponding surface wave, and Δ is the epicentral distance. If we use the T = 20 s, we calculate a magnitude of M S = 3.94 ± 0.18 with the ChinArray data and 4.03 ± 0.17 for the F-net data, slightly higher than the M S ~3.7 by Tauzin et al. (2013).
5 Discussion and conclusions
In this study, we conducted a systematic detection of weak seismic signal associated with the 2013 Russian meteor event. We identified clear seismic phases with BAZ of 329.7° and slowness of 34.73 s/deg on ChinArray with F-K analysis, consistent with them being the surface waves produced by the Russian meteor event. Combing with the best-fitting BAZ of 321.3° on the Japanese F-net, we located the Russian meteor event at 58.80°N, 58.72°E, which is about 438 km away from the location by the USGS (55.15°N, 61.41°E). The surface-wave phases from the M 5.8 Tonga earthquake ~20 min earlier mixed with the surface waves of the Russian meteor event. However, with vespagram and F-K analysis, we were able to identify phases of the Tonga earthquake with BAZ of 111.0° and slowness of 34.82 s/deg on the ChinArray, and with BAZ of 125.0° and slowness of 30.52 s/deg on the F-net.
One basic assumption of array processing techniques in the GAP software package is that the incident seismic waves are coherent plane waves (Koper 2005). ChinArray has a large aperture of ~800 km, which may not satisfy this assumption. This could explain the observation that the best-fitting BAZ of F-K analysis for the Russian meteor event becomes more robust if we use a small part of ChinArray (Fig. 4c). However, there is still a discrepancy with the true event location even when we use the stations within a small subarray (Figs. 4c, 11b, 12). As discussed before, the nonhomogeneity of the subsurface velocity structure may bend the propagation path of surface wave. For example, Lebedev and van der Hilst (2008) found a low-velocity anomaly of S wave around the boundary of Eurasian and Pacific Plate at the depth of 80 km, where the F-net is located. In addition, surface waves of the Russian meteor event passed through the Tianshan orogenic belt, where a relatively fast shear-wave speed at the depth of 75 km has been observed (Sun et al. 2010). The existence of these velocity anomalies may result in a larger discrepancy of the BAZ on F-net (~4.69°) than that on ChinArray (~3.73°) for the Russian meteor event, and a larger discrepancy of BAZ on ChinArray for the Russian meteor event (~3.73°) than that for the Tonga earthquake (~0.35°).
When performing the F-K analysis, we chose a higher filter band (0.05–0.07 vs. 0.03–0.05 Hz) and a shorter time window (100 vs. 140 s) for the Russian meteor event than the Tonga earthquake. Because the Russian meteor event has a shorter epicentral distance (~4700 vs. ~10,000 km to ChinArray), higher frequency energy could be observed. In addition, the Russian meteor event has a smaller magnitude and a shorter duration of the source time function right before the surface waves from the Tonga event. Hence, a shorter time window to extract the phase could avoid contamination from later arrivals.
Tauzin et al. (2013) relocated the Russian meteor event source with stations of GSN and FDSN within 4000 km away from the meteor. Seleznev et al. (2014) determined the exact time of the meteor explosion using West Siberian stations within 2000 km away from the meteor. The waveforms from ChinArray have longer epicentral distances of 4100–5200 km. Due to the dense distribution of stations within ChinArray, we could select part of the waveforms to reduce the interference from the Tonga earthquake, and detect the Russian meteor event with sub-ChinArray at such distances. On the other hand, with the best-fitting BAZ and slowness, we could stack all the waveforms within ChinArray, enhance the signals from the weak Russian meteor event significantly, and identify them out of those from the stronger Tonga earthquake. Although we were unable to accurately locate their source locations based on simple back-projection of the ChinArray and F-net results, our results shown in this study clearly demonstrate the capabilities of detecting surface waves of M ~4 events at long-range distances. Further studies are needed to evaluate its detection abilities of teleseismic P waves at higher frequency ranges.
- Borovička J, Spurný P, Brown P, Wiegert P, Kalenda P, Clark D, Shrbený L (2013) The trajectory, structure and origin of the Chelyabinsk asteroidal impactor. Nature 503(7475):235–237Google Scholar
- Davies D, Kelly EJ, Filson JR (1971) Vespa process for analysis of seismic signals. Nature 232:8–13Google Scholar
- Ding ZF, Wu ZL (2013). Advances of ChinArray Program. AGU, fall meeting 2013 abstract #S54A-01Google Scholar
- Koper KD (2005) The generic array processing (GAP) software package. Seismol Res Lett 76:226Google Scholar
- Mykkeltveit S, Ringdal F, Kværna T, Alewine RW (1990) Application of regional arrays in seismic verification research. Bull Seismol Soc Am 80(6B):1777–1800Google Scholar
- Ringdal F, Husebye ES (1982) Application of arrays in the detection, location, and identification of seismic events. Bull Seismol Soc Am 72(6B):S201–S224Google Scholar
- Wang RJ (1999) A simple orthonormalization method for stable and efficient computation of Green’s functions. Bull Seismol Soc Am 89(3):733–741Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.