The 2013 Chelyabinsk meteor ionospheric impact studied using GPS measurements



On 15 February 2013, the Chelyabinsk meteor event (the largest in size since 1908) provided a unique opportunity to observe ionospheric perturbations associated with the ablation and ionospheric impact of the meteor using GPS measurements. The hypersonic bolide generated powerful shock waves while acoustic perturbations in the atmosphere led to the upward propagation of acoustic and gravity waves into the ionosphere. In our research, we applied two different techniques to detect ionospheric disturbances in dual-frequency global positioning system (GPS) measurements during the meteor impact event. The data were collected from near-field GPS networks in Russia, GPS Earth Observation Network (GEONET) in Japan, and Plate Boundary Observatory (PBO) stations in the coterminous U.S. Using a novel wavelet coherence detection technique, we were able to identify three different wave trains in the measurements collected from the nearest GPS station to the meteor impact site, with frequencies of approximately 4.0–7.8 mHz, 1.0 −2.5 mHz, and 2.7–11 mHz at 03:30 UTC. We estimated the speed and direction of arrival of the total electron content (TEC) disturbances by cross-correlating TEC time series for every pair of stations in several areas of the GEONET and PBO networks. The results may be characterized as three different types of traveling ionospheric disturbances (TIDs). First, the higher-frequency (4.0–7.8 mHz) disturbances were observed around the station ARTU in Arti, Russia (56.43°N, 58.56°E), with an estimated mean propagation speed of about 862 ± 65 m/s (with 95% confidence interval). Another type of TID disturbance related to the wave trains was identified in the lower frequency band (1.0–2.5 mHz), propagating with a mean speed of 362 ± 23 m/s. The lower frequency ionospheric perturbations were observed at distances of 300–1500 km away from Chelyabinsk. The third type of TID wave train was identified using the PBO stations in the relative short-period range of 1.5–6 min (2.7–11 mHz) with a mean propagation speed of 733 ± 36 m/s. The observed short-period ionospheric perturbations in the U.S. region is, to the best of our knowledge, the first observational evidence of the coincident the long-range meteor-generated infrasound signals propagating in the ionosphere.

1 Introduction

A recent natural hazard event of high interest—the Chelyabinsk asteroid and resulting meteor, the largest since 1908—entered the Earth's atmosphere with an estimated speed of approximately 18.6 km/s and impacted Chelyabinsk, Russia, on 15 February 2013 [Borovička et al., 2013; Popova et al., 2013]. The estimated effective diameter of the asteroid was about 20 m, weighing about 10,000 metric tonnes, and the estimated total kinetic energy before the atmospheric impact was equivalent to 410 kt of Trinitrotoluene (TNT). Due to the shallow atmospheric entry angle of the asteroid, the Earth's atmosphere absorbed most of the energy from the generated explosions, shock waves, and heat. More specifically, the asteroid entered the atmosphere at 03:20 UT on 15 February 2013. The object subsequently burst into 11 individual pieces at a height of 39.2–29.8 km above the ground [e.g., Borovička et al., 2013]. Large fragments moving at high speed caused a powerful flash and strong shock waves, with most of the meteor's energy released at a height of 5–15 km above the Earth [Zuluaga et al., 2013]. Meteors deposit nanometer-sized smoke and dust particles in the Earth's upper atmosphere and ionosphere, triggering microphysical processes that affect the local ionization [McNeil et al., 2001]. They also generate infrasonic perturbations that can interact with the neutral atmosphere generating acoustic-gravity waves propagating into the ionosphere as measured by transionospheric Global Positioning System (GPS) and infrasound sensors [Le Pichon et al., 2013; Yang et al., 2013].

Ionospheric perturbations induced by acoustic and gravity waves generated in the neutral atmosphere are observed in GPS measurements. Events on Earth's surface or in the atmosphere, such as earthquakes, tsunamis, asteroid impacts, space shuttle launches, and large explosions are potential sources of ionospheric disturbances. Advancements in very high precision ionospheric GPS data processing have demonstrated that ground-based GPS receivers are capable of detecting total electron content (TEC) perturbations generated by atmospheric acoustic and gravity waves [Komjathy et al., 2012]. GPS data from the 2011 Tohoku earthquake and tsunami have, for instance, demonstrated that TEC perturbations due to gravity-wave activity are detectable within 45 min of earthquake onset [Galvan et al., 2012]. The 2004 Mount Asama volcanic eruption in Japan caused sinusoidal TEC perturbations with approximately 1 min period, and the earthquake near Samoa in September 2009 produced TEC fluctuations with an 8 min period [Galvan et al., 2011]. Additionally, a wavelet-based detection and estimation method was used to characterize various types of acoustic-gravity-wave-derived traveling ionospheric disturbances (TIDs) induced by tsunamis, earthquakes, and underground nuclear tests using GPS signals [Yang et al., 2011, 2012]. There remains much to learn about the characteristics of these interactions between the Earth's surface and ionosphere, including how and why they differ among different events.

In the research reported in this paper, two different techniques are applied to detect and analyze TIDs in the near field (Figure 1, Region 1). Method 1 (introduced in section 2) is used to generate the ionospheric perturbations in space and time, and a wavelet-based technique (Method 2) is used to mainly focus on the physical characteristics of the TIDs. The Chelyabinsk meteor impact has created a unique research environment that allows us to investigate for the first time how meteors affect the Earth's ionosphere using GPS measurements. Our goal is to isolate and analyze ionospheric responses associated with the meteor event, using measurements from GPS reference networks, as they could play a critical role in understanding the physics of the ionosphere's response to such sudden natural hazards. In our research, data were collected from stations near the impact location (Chelyabinsk, Russia, as shown in Figure 1), as well as sites located along the estimated asteroid trajectory (Japan, and western U.S.) within a 3 day window, centered on the event date. Section 2 describes the two techniques we used to analyze our data sets, while section 3 introduces the GPS data sets processed. Results are discussed in section 4 with a summary and concluding remarks given in section 5.

Figure 1.

Geometric relationship between GPS stations utilized (red dots), the reconstructed meteor trajectory in blue (with red lines as estimated error bounds for trajectory), the meteor impact location (red asterisk), and the subnetworks (blue boxes) for the TID estimation. The trajectory begins at the estimated atmospheric entry point.

2 Methodology

We applied two independent techniques to detect TEC perturbations in the vicinity of the meteor ground impact location including (1) Jet Propulsion Laboratory's (JPL's) PyIono algorithm [Komjathy et al., 2012] and (2) a wavelet-based detection and analysis scheme [Yang et al., 2012]. We imaged TEC perturbations induced by the asteroid atmospheric impact using GPS stations located in the vicinity of the impact region using Method 1. Subsequently, in Method 2, we applied a more sophisticated wavelet-based method and analyzed the TEC time series for identifying prominent coherent structures using three different regions. Lastly, we estimated the speed and direction of the meteor-generated TIDs using Method 2.

2.1 JPL PyIono Algorithm

In this method, we generated high-precision calibrated (also known as absolute) TEC measurements. Calibrating TEC measurements serves multiple purposes including quality checking of the processed data, leveling the phase measurements using pseudoranges, and comparing modeled and measured TEC perturbations [Mannucci et al., 1998]. Obtaining calibrated TEC values is important in understanding the background conditions for the perturbations [e.g., Komjathy et al., 2005]. We are primarily interested in monitoring short-term variations in the ionospheric electron content by analyzing changes in TEC measurements.

As a first step, JPL's PyIono algorithm uses TEC observations from dual-frequency code and phase measurements [e.g., Komjathy et al., 2012]. Subsequently, a Butterworth band-pass filter (corresponding to waves with periods between 33 and 3.3 min or frequencies between 0.5 and 5 mHz) is applied to isolate acoustic and gravity-wave-generated TEC disturbances. This type of filtering allows us to more easily detect perturbations within an expected range of frequencies corresponding to the acoustic- and gravity-wave-induced TIDs, based on past observations of multiple tsunami events [e.g., Galvan et al., 2012; Komjathy et al., 2012].

2.2 Wavelet-Based Detection and Estimation

A wavelet-based detection and estimation method for studying ionospheric disturbances induced by atmospheric acoustic-gravity waves was introduced by Yang et al. [2011]. A complex-valued continuous wavelet transform (CWT) with a Morlet wavelet math formula is applied to transform the TEC time series into time-frequency space, and compute the cross-wavelet spectrum using a wavelet coherence analysis by Yang et al. [2012], where ω0 (unitless) adjusts the scale resolution and η (unitless) is a nondimensional parameter [Mallat, 1999; Torrence and Compo, 1998]. A coherence analysis represents the degree of coherence of detected TEC perturbation structures (between 0 and 1) locally in the time and frequency domains. The detected structures are subsequently filtered, tuned to the frequencies at which disturbances are present [Yang et al., 2011]. This method analyzes TEC signals in a time-frequency space where disturbance signals may be classified and isolated based on their coherent wavelet spectrum as opposed to considering the disturbances only in one domain. A large and dense GPS network is first partitioned into several subnetworks (see examples identified with squares in Figure 1). To simplify the detection and estimation processes, we assume that TIDs propagate as planar waves through a small section defined as a 1° × 1° area within the region of a GPS network distribution. Subsequently, a continuous wavelet transform (CWT) is applied to convert TEC time series to time-frequency space with wavelet coefficients serving as inputs to the wavelet-based detection algorithm [Daubechies, 1990; El-Sheimy et al., 2003]. The next step is to utilize wavelet coherence techniques to detect and identify the GPS-derived ionospheric TEC disturbances in time-frequency space. In each small subnetwork, a threshold is set on the wavelet coherence spectrum to extract the wavelet coefficients corresponding to highly coherent structures, which are subsequently utilized for reconstructing the disturbance signatures as described in detail by Yang [2013]. Finally, a cross-correlation technique [Garrison et al., 2007] is applied to generate the reconstructed (or filtered) TEC time series corresponding to the classified disturbances in order to estimate propagation speeds [Yang et al., 2011, 2012]. In our research, the estimated propagation speed is represented with the expression math formula using 95% confidence limits, where math formula is the mean of the estimated speed, math formula is defined as the standard error of the mean, s is the sample standard deviation, and n as the size of the sample [Deep, 2006].

3 Description of Data Sets

GPS measurements used for this research were obtained from stations in the Japanese GEONET GPS network [Sagiya, 2004], the Plate Boundary Observatory (PBO) network (, and near-field GPS networks ( in the vicinity of the meteor ground impact location in Chelyabinsk. Three days' worth of data, centered on the event date, were used for data processing and analysis. By partitioning a full network into smaller subnetworks, it is feasible to map the variability of propagation speed and direction over a large area, while maintaining the assumption of a planar wave over each small subnetwork. Based on dual-frequency GPS data availability and global distribution of stations, we defined three regions (Figure 1, Regions 1 to 3) to investigate the ionospheric impact caused by the meteor. Chelyabinsk and its vicinity were designated as Region 1 (23 stations). The PBO stations on the west coast of the U.S. (Region 2–440 stations) and the Japanese GEONET network (Region 3–1235 stations) were divided into 32 and 66 subnetworks corresponding to each selected subarea (approximately 1° × 1° grids as shown in Figure 1), respectively.

4 Results and Discussion

In this section, we first examine the background ionospheric conditions. Subsequently, we summarize observational evidence of ionospheric responses coincident with the 2013 Chelyabinsk asteroid impact event based on GPS networks in three different regions as shown in Figure 1.

4.1 Background Ionospheric Conditions

As space weather events and geomagnetic storms can also result in ionospheric perturbations, we first investigated the global space weather indicators for 15 February 2013, as well as the surrounding days to understand their potential impact on our observations. Figure 2 presents the geomagnetic Kp indices from the NOAA/Space Weather Prediction Center. Moderate Kp index warnings were issued on 13 and 14 February 2013. However, it appears that there was no noteworthy geomagnetic event detected on 15 February 2013. Moreover, space and time TEC images in Figure 3 reveal quiet ionospheric conditions (indicated by green color and discussed in section 4.1) without strong ionospheric perturbations a day before and after the meteor impact. Fortunately, quiet ionospheric background conditions provide a favorable opportunity to distinguish between ionospheric disturbances that are due to the Chelyabinsk meteor ground impact event from those that are unrelated.

Figure 2.

Geomagnetic Kp indices (NOAA/Space Weather Prediction Center). Only moderate K index warnings were issued on 13–16 February 2013.

Figure 3.

The maps present TEC perturbations from a 3 day data set centered on the date of the Chelyabinsk asteroid event. The plots present all TEC disturbance time series corresponding to the ionospheric pierce points (IPP) for all receiver to satellite links using 23 stations in the near field. The approximate meteor trajectory is presented with a red line. The blue dashed lines are the estimated error bounds, and the meteor impact location is marked with a purple star.

4.2 Observed Ionospheric Responses Near Chelyabinsk

Stations in Region 1 (Figure 1) were processed using the two independent techniques. The results, obtained using Method 1 (band-pass filtering), suggest that individual explosions may have strongly affected the ionosphere. Figure 3 indicates TEC perturbations from the 23 GPS stations in Region 1 on the day before (Figures 3a and 3b), the day of (Figures 3c and 3d), and the day after (Figures 3e and 3f) the Chelyabinsk impact. The plots present TEC perturbations at the ionospheric pierce point (IPP) locations for all receiver-to-satellite links using GPS stations in Region 1. The results seem to indicate significant TEC perturbations (approximately at the 0.4 total electron content unit level, 1 TECU = 1016 el m−2) sustained for a few hours after the asteroid atmospheric impact, while the preceding and following days exhibit comparatively fewer significant variations.

The wavelet-based results seem to be in good agreement with the observations obtained from Method 1. Figure 4 illustrates the near-field coherence analysis result of wavelet detection using Method 2 (wavelets). The coherence results displayed in Figure 4 may be identified with two dominant frequency bands at 03:30 UT: 4.0–7.8 mHz and 1.0–2.5 mHz (with 95% confidence interval). Note that the frequency (on the vertical axis of Figures 4 and 6-8) is defined and plotted in logarithmic scale with base 2. The occurrence time of TEC signatures appears to be consistent with the ionospheric perturbations (in the first 2 h after the meteor impact) obtained with Method 1 using a Butterworth band-pass filter. Strong and coherent TEC perturbations were detected after the asteroid impact with the Earth's atmosphere and subsequently observed by GPS measurements. The coherent TEC disturbance signals were extracted to estimate propagation characteristics such as speed and direction. Figure 5 presents results of the two different (long- and short-period) TEC disturbances reconstructed for two independent station-to-satellite links (ARTU-PRN18 and ARTU-PRN26) by the wavelet coefficients within the two highly coherent frequency bands indicated in Figure 4. Furthermore, there seems to be no significant coherent structures detected 1 day before or after the event date as shown in Figures 6 and 7.

Figure 4.

Coherent spectrum of TEC measurements using station ARTU. Solid black contours represent the 95% confidence interval. Regions outside the black dashed line are affected by the edge effect due to the finite number of TEC data points. The vertical axis is defined and plotted in a logarithmic scale of base 2. The two dominant frequency bands are 4.0–7.8 mHz and 1.0–2.5 mHz at 03:30 UT. Color scale represents the coherence quantities (between 0 and 1).

Figure 5.

Long-period (blue lines) and short-period (red lines) TEC disturbance measurements reconstructed using the highly coherent wavelet coefficients within the two dominant regions in Figure 4 for links (a, c) ARTU-PRN18 and (b, d) ARTU-PRN26 on 15 February 2013.

Figure 6.

Coherent spectrum of TEC time series from station ARTU on 14 February 2013 (1 day before the Chelyabinsk event). No significant coherent structures are detected. See Figure 4 for further details.

Figure 7.

Coherent spectrum of TEC time series from station ARTU on 16 February 2013 (1 day after the Chelyabinsk event). No coherent structures are seen. See Figure 4 for further details.

On the other hand, coherence detection results using measurements of a near-field GPS network (23 stations approximately 1500 km south of Chelyabinsk) only identify the lower frequency structures in the TEC time series, as shown in Figure 8. The propagation speed, estimated using Method 2, is 362 ± 23 m/s.

Figure 8.

Coherent spectrum of TEC time series from measurements of a near-field GPS network in the region 1500 km south of Chelyabinsk. The coherence detection is only able to identify the lower frequency structures.

Based on the results using coherence detection, the speed and direction of the disturbances' arrival may be estimated by cross-correlating data points from all pairs of GPS stations in several subareas of each network [e.g., Yang et al., 2011]. They may be characterized as two different groups of TIDs, which are distinguished from their wavelet coherence detections as demonstrated in Figure 4. In the near field, the higher-frequency (4.0–7.8 mHz) disturbances with an estimated propagation speed of 862 ± 65 m/s (with 95% confidence) were observed within 300 km of the meteor impact with the ground. The second TID disturbance wave train was identified with a lower frequency band (1.0–2.5 mHz) propagating with a speed of 362 ± 23 m/s. The lower frequency TEC perturbations (the second wave train) were observed between 200 km and 1500 km away from Chelyabinsk.

4.3 Ionospheric Perturbations Observed in the U.S. Region

From the GPS data collected in Region 2, we observed short-period TIDs (2–6 min) with propagation speeds of 733 ± 36 m/s (95% confidence interval) in a near westerly propagation direction of 268.5 ± 5.0 degrees (95% confidence interval; 0° is north) estimated using the PBO network (420 stations). Significantly, strong short-period TIDs were observed about 3 h and 36 min after the meteor impact (at 03:20 UTC). They appeared to propagate from the east to the west coast of the U.S. and were sustained for about 30 min. Their propagation speeds seem to be consistent with infrasound speeds in the ionosphere (at a height of 300 km) [Le Pichon et al., 2009]. The estimated vectors in Figure 9 represent all detected TIDs from each subarea in Region 2, and the color-scaled points represent the filtered TEC amplitudes at each IPP corresponding to each line of sight pair using PBO GPS stations to PRN 8 and PRN 17. In addition, we calculated the approximate average global travel speed defined as the great circle distance from the meteor impact location to the IPP position corresponding to the strongest TEC perturbations divided by the travel time (3 h and 36 min). The average global speed is calculated to be approximately 756 m/s. The speed seems to be in good agreement with our estimates using PBO stations in the U.S. region. This indicates that the detected TIDs may be caused by the asteroid atmospheric impact.

Figure 9.

Snapshot of TEC perturbations following the asteroid atmospheric impact in Region 2 at 06:56 UTC. Color-scaled points represent the TEC perturbation magnitudes (Method 2) at each ionospheric pierce point (in geodetic longitude and latitude) corresponding to each line of sight station-to-satellite pair for the coterminous GPS stations to PRN 8 and PRN 17. Blue arrows and the black vector represent the estimated TID propagation velocities and a reference velocity vector, respectively. The starting points of the blue arrows represent the IPP locations with maximum TEC perturbations in a subarea [Yang, 2013]. Strong wavelike disturbance structures were observed 3 h and 36 min after the meteor ground impact.

Moreover, recent results by Le Pichon et al. [2013] concluded that the far-field infrasound signals associated with the Chelyabinsk asteroid atmospheric impact were observed at multiple locations globally using the Comprehensive Nuclear-Test-Ban Treaty Organization's International Monitoring System Infrasound Network ( Specifically, the long-range infrasound signals took approximately 10 h to travel from the impact location to infrasound station I57US in California [Le Pichon et al., 2013]. As a comparison, we observed that a corresponding travel time for TIDs to reach GPS stations in California is about 3 h and 36 min. A preliminary analysis suggests that the observed travel times seem to be consistent with the propagation speed differences of acoustic waves in the lower troposphere (approximately 334 m/s) compared to the ionosphere (700–900 m/s in the F region). An in-depth analysis of propagation properties of the acoustic waves using modeling and ray-tracing techniques is still underway.

4.4 Processing Results From Japan

Figure 10 presents snapshots of TEC images using 1235 GEONET stations observing GPS satellite PRN5 on the event day of 15 February 2013 (Figures 10a and 10b), a day before (Figure 10c), and a day after (Figure 10d). Strong turbulence-like disturbance structures are observed 26 min after the meteor ground impact as shown in Figure 10a. The phenomena were sustained for more than 1 h (Figure 10b). On the day before (14 February; see Figure 10c), we observed nonwavelike TEC signatures and they appear to be localized to the northern and southern parts of Japan. No strong signals were detected on the day after (16 February) the meteor ground impact event. On the event day, the turbulence-like disturbances appear to be distinctly different from the wavelike TID observations in other regions. Furthermore, a hypersonic meteor entry typically generates shockwaves, infrasound, and acoustic-gravity waves perpendicular to the meteor trajectory propagating out from the Mach cone [Le Pichon et al., 2009]. Therefore, the absence of the wavelike TIDs seems to be in agreement with the classical meteor shock wave and infrasound theory [Le Pichon et al., 2009].

Figure 10.

Snapshots of TEC perturbations on the meteor (a, b) impact day, (c) a day before, and (d) a day after using 1235 GEONET stations.

5 Conclusions

Traveling ionospheric disturbances, resulting from the Chelyabinsk meteor ground impact in Russia have been detected and identified and their temporal and spatial structure analyzed using GPS measurements with two different methods. Method 1 was used for imaging the resulting TEC perturbations, while Method 2 was applied to identify coherent structures and estimate speed and direction of TIDs. Results of monitoring TEC perturbations with the two techniques suggest a strong impact of the individual explosions on the ionosphere. Furthermore, we characterized the observed TEC perturbations and grouped them into three different wave trains.

First, at station ARTU, higher-frequency (4.0–7.8 mHz) perturbations were detected with an estimated mean propagation speed of about 862 ± 65 m/s (95% confidence interval). Second, in the near field, a second type of TID was observed with a frequency band of 1.0–2.5 mHz with a mean speed of 362 ± 23 m/s. Third, in the far field, the third type of TID was discovered with a shorter period of 1.5–6 min corresponding to 2.7–11 mHz and mean speed of 733 ± 36 m/s.

The slower speed disturbances seem to be consistent with the TIDs induced by gravity waves as presented in previous studies [e.g., Galvan et al., 2012; Komjathy et al., 2005; Hickey et al., 2009; Yang et al., 2012]. However, to the best of our knowledge, the short-period disturbance waves with higher propagation speeds (the third type) have previously not been identified for other natural hazards such as tsunamis, earthquakes, or volcano eruptions and surface explosions. Due to the complexity of atmospheric and ionospheric physics associated with meteor ablation, GPS-derived ionospheric TEC perturbations may provide additional clues for better understanding the interactions between the predominant neutral species in the ionosphere and acoustic- gravity waves generated by asteroid atmospheric impacts.

In addition, the detection of short-period TIDs observed in the U.S. region implies that they are associated with the meteor acoustic waves (infrasound signals). Their periods, propagation speeds, and directions seem to be in agreement with the Chelyabinsk asteroid atmospheric impact as the origin. We understand this is the first observational evidence regarding the ionospheric perturbations coincident with the long-range meteor-generated infrasound signals propagating in the ionosphere, which are different from the infrasound signals ducted between the thermosphere and Earth's ground discussed by Le Pichon et al. [2013]. Additional work is needed to verify if the observed TIDs may be associated with specific physical processes of the meteor ground impact on the ionosphere.


The authors would like to thank NASA Headquarters, the NASA ROSES 2011 GNSS Remote Sensing Science Team (NNH11ZDA001N-GNSS), and the NASA Postdoctoral Program (NPP) administrated by Oak Ridge Associated Universities. We would also like to thank the anonymous reviewers of our manuscript for helpful comments. Our research was performed at the Jet Propulsion Laboratory, California Institute of Technology, under contract to the National Aeronautics and Space Administration.