On 15 February 2013, a large Earth-impacting fireball disintegrated over the Ural Mountains. This extraordinary event is, together with the 1908 Tunguska fireball, among the most energetic events ever instrumentally recorded. It generated infrasound returns, after circling the globe, at distances up to ~85,000 km, and was detected at 20 infrasonic stations of the global International Monitoring System (IMS). For the first time since the establishment of the IMS infrasound network, multiple arrivals involving waves that traveled twice round the globe have been clearly identified. A preliminary estimate of the explosive energy using empirical period-yield scaling relations gives a value of 460 kt of TNT equivalent. In the context of the future verification of the Comprehensive Nuclear-Test-Ban Treaty, this event provides a prominent milestone for studying in detail infrasound propagation around the globe for almost 3 days as well as for calibrating the performance of the IMS network.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 At 03:20:26 UTC on 15 February 2013, a large Earth-impacting fireball entered the Earth's atmosphere over the Kazakhstan/Russia border. According to the National Aeronautics and Space Administration (NASA) agency (report available at http://www.nasa.gov/mission pages/asteroids/news/asteroid20130215.html), the object was a relatively small asteroid that entered the Earth's atmosphere at high speed and at a shallow angle. The approximate effective diameter of the asteroid when it penetrated the atmosphere was estimated to be 17 m and its mass about 10,000 t. Traveling northwest over Russia at a velocity of about 20 km/s, it reached its maximum brightness south of Chelyabinsk, Russia (54.80°N, 61.10°E), at an altitude near 25 km (report from the Department of Physics and Astronomy, University of Western Ontario, Canada, available at http://physics.uwo.ca/news/chelyabinsk_meteor_15feb2013_data1.html). During the atmospheric entry phase, the hypersonic impacting object fragmented when heated by atmospheric friction. The shock waves generated by this event blew out windows and damaged buildings in several surrounding cities. The energy released by the burning mass was large enough to generate low-frequency pressure waves which have been detected worldwide. This event is exceptional in terms of the number of remote infrasound stations that detected signals and the very long distances of propagation.
 The global International Monitoring System (IMS) infrasound network operated by the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO) has been designed to detect any violations of the treaty [Christie and Campus, 2010]. Although not yet fully established, it has demonstrated its capability to detect and locate geophysical and anthropogenic sources on a global scale [e.g., Arrowsmith et al., 2008; Campus and Christie, 2010; Hedlin and Walker, 2012]. The IMS network detected worldwide infrasonic waves generated by the blast of the 2013 Russian event (press release available at http://www.ctbto.org/press-centre/press-releases/2013/russian-fireball-largest-ever-detected-by-ctbtos infrasound-sensors and http://media.egu2013.eu/press-conference/#meteor). This fireball is the most energetic event reported since 1908, when a meteor broke up over Siberia's Tunguska River. It is also the largest event ever registered by the CTBTO network [Schiermeier, 2013; Showstack, 2013]. The 8 October 2009 meteor that exploded over Indonesia is the most recent similar event detected by 17 IMS stations with an explosive energy estimated at about 50 kt of TNT equivalent [Silber et al., 2011].
 The hypersonic entry of meteoroids generates infrasonic waves which are refracted and channeled over long distances by the temperature gradient and the wind structure of the atmosphere [Kulichkov, 1992]. Meteoroid explosions can be observed a few thousand kilometers away from the explosion point when they penetrate the atmosphere below 60 km altitude [ReVelle, 1997; Brown et al., 2002b; Edwards, 2010]. Large supersonic objects penetrate deeper into the atmosphere, sometimes impacting the ground, and may be modeled as a line source explosion. Although very large objects, like the 1908 Tunguska meteor, can generate low-frequency gravity waves, most of them excite acoustic waves in the infrasound domain [ReVelle et al., 2008]. Global infrasound measurements of fireball disintegration represent valuable input data for estimating and validating the influx rate of meter-sized and larger meteoroids [Brown et al., 2002c]. They can also provide crucial information about trajectory and energy for events of interest which otherwise lack such information [e.g., Brown et al., 2008; Le Pichon et al., 2008; Chapman, 2008].
 Here we present a global analysis of the signals recorded by the IMS infrasound network. In the first part, we detail the detection procedure which has been used to extract the wave parameters of the signals. In the second part, long-range propagation features are analyzed and interpreted to give an accurate description of atmospheric dynamics on a global scale. Finally, we provide a rough estimate of the total explosive energy released by the fireball and discuss the importance of such recordings to survey and identify hazardous near-Earth objects entering the atmosphere which may collide with the Earth.
2 Worldwide Infrasonic Detections
 The International Data Centre (IDC) of the CTBTO in Vienna processes automatically and in near real time continuous recordings from the globally deployed IMS infrasound stations. The IDC automatic system is designed to detect explosion-like signals. Station detections are associated to form events. The system can associate signal detections at distances up to 6700 km from the source location. The Russian fireball was automatically detected by the IDC automatic system. The reviewed analysis carried out in the hours following the event provided an extended list of infrasound signals associated with the meteor as well as a refined source location.
 In the Reviewed Event Bulletin of the IDC, signals recorded at 19 of the 42 operational infrasound stations and four seismic stations were associated to build an event near the Ural Mountains regions, 130 km south of Chelyabinsk, at coordinates 54.06°N, 61.81°E with an error ellipse of ~1° (major axis). The source origin time of the main blast is estimated at 03:22:06 UTC with an error in excess of 200 s. These results coincide with the location and time of maximum brightness of the incoming meteor.
 In order to better characterize the signal wave parameters in a broad frequency band, 4 days of continuous recordings following the occurrence of the event of the 42 operating IMS stations were systematically reprocessed using the progressive multichannel correlation (PMCC) method [Cansi, 1995]. PMCC was configured with 20 logarithmic-spaced frequency bands, adapted from Matoza et al. , with Chebyshev filters of order 2 between 0.01 and 4 Hz. Compared with the IDC automatic processing using 11 bands between 0.07 and 4 Hz, this logarithmic-spaced configuration permits computationally efficient broadband processing and helps with signal discrimination [Brachet et al., 2010]. We vary the time window length in proportion to the period from 30 to 200 s; the window is time shifted by 10% of the window length. Since we focus on signals with frequencies as low as 0.01 Hz, we choose subnetwork geometries with the largest array element separations. By applying this detection procedure, infrasound signals were found at 20 IMS stations.
 At most stations, substantial energy is found at periods larger than 20 s, which is consistent with high explosive energy. Of specific interest are arrivals involving waves that traveled along the minor and major great circle arcs (referred to as Ig1 and Ig2, respectively). For example, arrival Ig1 is observed at IS27 in Antarctica after a propagation over a distance of 14,948 km, while arrival Ig2 that traveled over a distance of 25,127 km is detected about 9 h later in the antipodal direction. Additional arrivals, Ig3 and Ig5, which circled once and twice round the globe are also clearly observed up to a maximum distance of 86,663 km. Overall, 14 propagation paths are identified as Ig2, Ig3, and Ig5. Such arrivals have not been observed since the last major eruption of Mount Saint Helens in 1980 [Delclos et al., 1990].
 Figure 1 presents an example of an Ig5 arrival at station IS53 in Alaska. The signals detected in the 0.02–0.05 Hz frequency band (pink color) are characterized by a coherent emergent waveform that decays slowly and lasts approximately 2 h. At such propagation range, the long signal duration is interpreted as the combination of multiple paths in a ground-to-stratosphere acoustic waveguide [e.g., Green et al., 2011]. In this time window, microbarom signals (blue color) resulting from nonlinear interacting ocean waves in northern Pacific regions dominate the band 0.08–0.5 Hz, with amplitude peaked at 0.2 Hz [Landès et al., 2012]. Table S1 in the supporting information summarizes the signal characteristics of the detected coherent waves that are consistent in azimuth and arrival time—assuming stratospheric or thermospheric returns—with the origin of the event. Signals are remarkably visible on the waveform despite the extreme ranges of detections (e.g., IS18 in Greenland and IS53 in Alaska).
 Table S1 also indicates the maximum of the effective sound speed ratio (Veff-ratio). This dimensionless parameter is a proxy for the combined effects of refraction due to sound speed gradients and along-path wind advection on infrasound propagation [e.g., Le Pichon et al., 2012]. It is defined as the ratio between the maximum of the along-path wind plus the adiabatic sound speed at 30–60 km altitude and the sound speed at the ground level, averaged along each propagation path.
 For most downwind stations (Veff-ratio ~ 1 or larger), stratospheric arrivals are expected with celerity values ranging from 280 to 320 m/s with a mean value of 294 m/s. Thermospheric returns observed as far as 43,340 km (Ig1 at IS43, Ig1 and Ig3 at IS26, and Ig1 at IS18) are noteworthy. The relatively low celerity values (267, 241, 277, and 280 m/s, respectively) are consistent with such phase identification for stations located upwind [Brown et al., 2002a].
 The main observational results are outlined below.
 At 530 km, the signal amplitude is 12 Pa and decreases down to ~0.1 Pa at distances larger than ~40,000 km. The largest amplitude observed downwind is 18 Pa at IS53 (Alaska), about 6000 km from the source. This amplitude seems to be anomalously large. Further studies are needed to check the instrument response and investigate possible focusing effects using full-wave propagation simulations.
 The dominant period ranges from ~20 to ~70 s with a mean value of 39.9 s. Such large variations in the observed periods might be explained by the combined effects of both signals arriving from different portions of the fireball trajectory and the spatiotemporal variations of the atmosphere over long propagation range [Edwards et al., 2006].
 For downwind stations, only a small attenuation is observed over large distances. For example, the amplitudes of the Ig3 and Ig5 signals at IS53 are almost the same (0.6 and 0.5 Pa, respectively) due to weak atmospheric absorption for signal frequency lower than 0.1 Hz. Considering downwind propagation (Veff-ratio > 1), the predicted attenuation of Ig5 compared with Ig3 is about 5 dB [Le Pichon et al., 2012].
 The long coda of the signals (up to ~3 h at IS57 for Ig5) may be due to the combined effects of the spatiotemporal extension of the source [e.g., Evers and Haak, 2001; Brown et al., 2003] and multiple reverberations in acoustic waveguides [e.g., Kulichkov, 2004; Ceranna et al., 2009].
3 Long-Range Infrasound Propagation
 The waves generated by the fireball's disintegration propagated over very long distances. Figure S1 presents a global picture of the stratospheric wind dynamics at 50 km altitude. On 15 February 2013 at 06:00 UTC, dominant winds were blowing eastward in the Northern Hemisphere at mid-latitude regions. These weather conditions remained relatively stable over the following 3 days with prevailing eastward stratospheric jet at mid-latitude in the Northern Hemisphere. Stations located to the east of the event were thus favorably positioned for detecting signals. Figure 2 presents barograms of the recordings, highlighting Ig1 and Ig2 arrivals at 20 IMS stations. At some stations (e.g., IS43, IS18, and IS53), signals are clearly visible above the background noise level, even at very remote sites (e.g., IS27).
 Using the European Centre for Medium-Range Weather Forecasts (ECMWF) wind and temperature specifications (http://www.ecmwf.int/), travel times for Ig1 and Ig2 waves are calculated in each direction with radial distance segmented every 5° (Figure S2). The effective sound speed ratio along each propagation direction is calculated and displayed. For each cell, travel times are computed using the tau-p approach [Garcés et al., 1998] focusing on stratospheric and thermospheric arrivals and assuming that due to the long period of the signals, tropospheric ducts are unlikely. At each cell, the median of the celerity values of the fastest predicted branch is calculated. Then, travel times toward an arbitrary point on the global grid are computed by summing all incremental distances along the great circle path originating from the source multiplied by the corresponding celerity values. Good agreements between the measured and predicted celerity values are obtained, except for Ig1 paths toward IS31 and IS43 (Table S1). These discrepancies can be explained by uncertainties in the description on the stratospheric waveguide structure and the known limitations of the high-frequency approximation of the ray tracing method used [Drob et al., 2010], as well as the relatively coarse gridding that is used in the calculations of celerity for these stations when the source is a line source.
 Usually, the highly variable atmospheric conditions in both space and time complicate the propagation modeling. However, an attempt to simulate long propagation range is performed here using a 2-D ray tracing program that is part of the Consortium for Research in Elastic Wave Exploration Seismology (CREWES) software package (http://www.crewes.org/Reports/2000/2000-09.pdf written by G. Margrave). Allowing limited pressure perturbations, the acoustic propagation is governed by the linearized equations for a compressible fluid. This implies that the signal wavelengths are smaller than those of atmospheric property variations. However, considering the wavelength of the propagating signals (of the order of 10 km), the high-frequency approximation of the ray tracing approach is still relevant to predict wide geometric acoustic ducts where infrasound energy can efficiently propagate. The atmospheric specifications are similar for several days following the date of the event with stable prevailing stratospheric wind jets blowing in opposite directions in both hemispheres. The temporal fluctuations of both temperature and horizontal wind terms during the propagation are therefore neglected. For the modeling, the global ECMWF-91 profiles (with a horizontal resolution of 0.5°) are merged at the top of the model (0.01 hPa) with the MSISE-00 and HWM-07 climatologies [Drob et al., 2008] in order to extend the profiles into the thermosphere. Wind and temperature profiles are interpolated over a spatial grid of 200 m × 200 m resolution in range and altitude, respectively. Moreover, a flat atmosphere transform [Müller, 1985] is applied to account for the Earth's curvature.
 At almost all stations, ray tracing results accounting for a range-dependent meteorology show that the acoustic energy efficiently propagated between the source and receivers in the lowermost 60 km of the atmosphere. Figure S3 presents examples of ray tracing simulations toward IS21 (Marquesas Islands), IS27 (Antarctica), and IS26 (Germany). Rays are chosen to best fit the observed celerity and the measured horizontal trace velocity. Paths toward IS26 are dominated by zonal atmospheric conditions, whereas the other two are dominated by meridional conditions. Figure S3a presents simulation results toward IS21 along both the minor and major great circle arcs (Ig1 and Ig2, respectively). For an explosive source at 25 km altitude, simulations predict in both directions energy being trapped in elevated stratospheric ducts between 10 and 40 km altitudes. The bottom of the waveguides above the station is 13 km for Ig1 and 7 km for Ig2. These simulations provide a good explanation of the Ig2 arrival which returned to the surface through diffraction or leakage out of the duct [e.g., Brown et al., 2003]. Figure S3b shows the predicted paths toward IS27. As for IS21, the returned energy for Ig1 and Ig2 can be explained by the simulated ray paths. The thermospheric to stratospheric phase conversion for Ig2 is explained by the fairly weak meridional wind at the source and a reinforced stratospheric duct which persists for most of the propagation. The observed amplitudes and duration show higher values for Ig2, which might be explained by the less elevated duct compared to Ig1. Figure S3c presents simulation results toward IS26 for Ig1 and Ig3. The westward propagation exhibits phase conversion from thermospheric in the Northern Hemisphere (upwind) to stratospheric in the Southern Hemisphere (downwind) and finally back to thermospheric across the equator. Due to stronger along-path wind at 30–60 km altitude, at range larger than ~5000 km, mesospheric and thermospheric paths were channeled into stratospheric ducts and refracted back to the ground at about 30 km altitude. Above the receiver, only rays propagating in an elevated thermospheric duct are predicted. These predictions are consistent with the low value of the measured celerity for Ig1 (241 m/s) and slightly higher for Ig3 (277 m/s). Such long propagation paths are explained by negligible absorption in the thermosphere at a frequency of 0.02–0.05 Hz [Sutherland and Bass, 2004].
4 Estimating the Source Energy
 During the hypersonic entry of a meteor into the atmosphere, the object is strongly decelerated and disintegrates. The explosive energy released in the atmosphere by fragmentation can be calculated using empirical relations based on a series of measurements obtained from low-altitude nuclear explosions. The result corresponds to the yield of an explosion in equivalent tons of TNT which would produce the same effects as the shock induced by the penetration of the near-Earth object into the atmosphere.
 There are several empirical relations relying on either the dominant signal period or amplitude, which can be used to infer the explosive yield from infrasound measurements [e.g., Edwards et al., 2006]. The frequency content of the signals detected at long propagation range is generally less modified than the signal amplitude; thus, the period relationship is expected to be more robust [Mutschlecner et al., 1999]. A commonly used empirical period-yield relation was developed by the U.S. Air Force Technical Applications Center. The regression to these data is given by ReVelle :
where W is the total source yield (in kilotons of equivalent TNT; 1 kt TNT = 4.185 × 1012 J) and Τ is the period (in seconds) of the observed infrasonic signal at maximum amplitude. Compared with other empirical relations, this approach has shown consistent agreement with yield estimates for bolide events when energies could be obtained with other methods [Brown et al., 2002a; Silber et al., 2011; Ens et al., 2012]. The mean period of signals recorded at all stations listed in Table S1 is 39.9 ± 10.3s. Considering all IMS stations at ranges less than 10,000 km, the application of equation ((1)) gives yield values ranging from 100 kt to 1.4 Mt, with a geometrical mean of 460 kt. Even though there is a large spread in this estimate, the obtained mean value is comparable with the total impact energy of 440 kt which has been derived from the measured optical radiant energy published by NASA (report available at http://neo.jpl.nasa.gov/fireballs/).
5 Concluding Remarks
 Infrasound signals produced by the blast of the 2013 Russian fireball were observed at extreme ranges by 20 IMS stations. By applying a logarithmic-frequency-scaled implementation of the PMCC method, signals which circled twice round the globe were detected with coherent energy at periods as large as 70 s. The multiple reverberations of the waves in the acoustic waveguides explain the long signal duration reaching nearly 3 h. At such low frequency, thermospheric signals were also observed upwind with very little attenuation over one full Earth's circumference. Of specific interest is the equatorial penetration of waves due to sufficient meridional winds which reinforce stratospheric ducting. Continuing in-depth analysis of these low-frequency recordings would likely extend the existing list of arrivals.
 The long period of the signals is consistent with the high source energy and large blast cavity. With a mean dominant period of 39.9 s, the calculated explosive energy is 460 kt with 1 order of magnitude uncertainty. This event is the most energetic one ever detected by the global IMS infrasound network. Based on the expected flux rate of small near-Earth objects colliding with the Earth [Toon et al., 1997; Chyba et al., 1998; Brown et al., 2002b], rocks with the size of the Russian fireball are expected to occur about once every 50–100 years. This event offers a rare opportunity to provide insight into the threshold levels at which ground damage may be caused by airbursts.
 Together with recent advances in modeling the atmosphere [e.g., Drob et al., 2013], future works using numerical methods with fully resolved, time- and range-dependent wave propagation techniques, beyond the scope of this study, would provide more detailed explanations of the characteristics of the received signals [Millet et al., 2007; Norris et al., 2010]. In particular, the used frequency-dependent attenuation relations in a realistic atmosphere could provide additional constraints on the source yield estimate [Le Pichon et al., 2012].
 In general, this event opens up avenues for evaluating station sites, because it was not measured at all IMS infrasound arrays, although at some stations, signals circulating twice round the globe were recorded. Moreover, combining updated broadband detection algorithms and range-dependent propagation modeling in a realistic atmosphere will improve knowledge of the physical mechanisms involved in near-Earth object interaction with the atmosphere and document possible local atmospheric perturbations.
 Finally, continuing such studies will provide a benchmark for future studies on exploding fireballs. These observations are especially useful when events occur during the daytime and records from other observing instruments, such as Earth-based telescopes and all-sky cameras, cannot be made reliably or when an event is located over a remote area such as the vast open oceans where no surface-based recording instruments are available. In addition, it is expected that more detailed analyses of such large events will result in the development of techniques to discriminate between exploding fireballs and atmospheric nuclear explosions and would help to advance the development of monitoring procedures to identify potentially dangerous exploding near-Earth objects.
 We are grateful to P. G. Brown (Department of Physics and Astronomy, University of Western Ontario, Canada) and W. N. Edwards (Canadian Hazard Information Service, Natural Resources Canada, Ontario, Canada) for helpful discussions. We wish to thank M. Hoffmann and T. Grasse for bringing our attention to recordings in Antarctica at station IS27; they were the first in finding signals propagated along the short and long great circle arcs. Many thanks also to Y. Cansi and J. Vergoz for the time they spent helping us in the data analysis during the completion of this work. This work was partly performed during the course of the ARISE collaborative project of the Seventh Framework Programme, funded by the European Union (http://arise-project.eu/).
 The Editor thanks Michael Hedlin and an anonymous reviewer for their assistance in evaluating this paper.