Statistical analysis of fireballs: Seismic signature survey
T. Neidhart, K. Miljkovi?, E.K. Sansom, H.A.R. Devillepoix, T. Kawamura, J.-L. Dimech, M.A. Wieczorek, P.A. Bland
PPublications of the Astronomical Society of Australia (PASA)doi: 10.1017/pas.2021.xxx.
Statistical analysis of fireballs: Seismic signature survey
T. Neidhart , K. Miljković , E.K. Sansom , H.A.R. Devillepoix , T. Kawamura , J.-L. Dimech ,M.A. Wieczorek , P.A. Bland School of Earth and Planetary Sciences, Space Science and Technology Centre, Curtin University, Perth, Australia Institut de Physique du Globe de Paris, France Geoscience Australia, Canberra, Australia Observatoire de Cote d’Azur, Laboratoire Lagrange, Nice, France
Abstract
Fireballs are infrequently recorded by seismic sensors on the ground. If recorded, they are usuallyreported as one-off events. This study is the first seismic bulk analysis of the largest single fireball dataset, observed by the Desert Fireball Network (DFN) in Australia in the period 2014–2019. The DFNtypically observes fireballs from cm-m scale impactors. We identified 25 fireballs in seismic time seriesdata recorded by the Australian National Seismograph Network (ANSN). This corresponds to 1.8%of surveyed fireballs, at the kinetic energy range of 10 to 10 J. The peaks observed in the seismictime series data were consistent with calculated arrival times of the direct airwave or ground-coupledRayleigh wave caused by shock waves by the fireball in the atmosphere (either due to fragmentation orthe passage of the Mach cone). Our work suggests that identification of fireball events in the seismic timeseries data depends both on physical properties of a fireball (such as fireball energy and entry angle inthe atmosphere) and the sensitivity of a seismic instrument. This work suggests that fireballs are likelydetectable within 200 km direct air distance between a fireball and seismic station, for sensors used inthe ANSN. If each DFN observatory had been accompanied by a seismic sensor of similar sensitivity,50% of surveyed fireballs could have been detected. These statistics justify the future consideration ofexpanding the DFN camera network into the seismic domain.
Keywords: fireball – impact – seismic – observation – sensitivity
When a meteoroid enters the atmosphere, it experi-ences aerodynamic drag and dynamic pressure. Theatmosphere slows down meteoroids and in most casesthey break-up and vaporize (Ceplecha & Revelle, 2005).The break-up occurs when the dynamic pressure ishigher than its compression strength (Cevolani, 1994;Stevanović et al., 2017). Shock waves can be generatedin the atmosphere by (Figure 1):• The hypersonic flight forming a Mach cone,• A discrete fragmentation event during the mete-oroid’s trajectory,• A catastrophic final airburst,• Physical impact on the ground (extremely rare).The Mach angle within the Mach cone is expected tobe negligibly small, because the impact speed is muchlarger than the speed of sound in the air. Therefore,the shock waves generated during a hypersonic fireballentry are expected to propagate almost perpendicularto the trajectory (Figure 1a). The fragmentation of the meteoroid can also create shock waves that propagatewith no preferred direction; thus, can be assumed theypropagate omnidirectionally (Figure 1b). If the impactoror parts of the impactor survive the atmospheric pathand hit the ground (Figure 1c), the seismic waves in theground can be generated by the impact itself (Edwardset al., 2008; Tancredi et al., 2009). The atmosphericshock waves can couple with the ground and form bodyand surface waves (Figure 1d) (Brown et al., 2003; Ste-vanović et al., 2017; Karakostas et al., 2018). The arrivaltimes for different seismic waves differ as they travelat different speeds through different media (ground orair), which allows for their classification. Airwaves gen-erated by the Mach cone (Figure 1e) will arrive last asthey travel slowest (at the speed of sound), through theair directly between the fireball and the sensor on theground (Edwards et al., 2008).For larger (bolide and cratering) events, a varietyof seismic waves has been recorded. For example, theseismic signals caused by the 20-m diameter asteroidthat exploded over Chelyabinsk, Russia in 2013 (esti-1 a r X i v : . [ a s t r o - ph . E P ] F e b Neidhart et al.
Figure 1.
Shock wave generation during a fireball event: (a)Shock waves are generated by the Mach cone that travel almostperpendicular to the trajectory of the object and rapidly decayfrom a non-linear to linear wavefront, (b) fragmentation-inducedairburst causes shock waves that travel omnidirectionally, (c)seismic waves originating from impact itself (d) Rayleigh wavesformed by coupling between airwaves and the ground, and (e) anair disturbance directed at the seismic station (Brown et al., 2003;Revelle et al., 2004). Figure redrawn from Edwards et al. (2008). mated to have carried 10 J at airburst (Emel’yanenkoet al., 2013)) were identified as P and S body waves,ground-coupled airwaves and Rayleigh waves (Tauzinet al., 2013); The P and S seismic waves were also seenwhen the 13.5-m diameter crater formed near Caran-cas, Peru in 2007 (Brown et al., 2008; Le Pichon et al.,2008; Tancredi et al., 2009); The Neuschwanstein largemeteorite (estimated to have had 10 J initial sourceenergy) (Revelle et al., 2004; Oberst et al., 2004) causedseismic activity by direct airwaves and ground coupledRayleigh waves at seismic stations within a few hundredkm distance (Revelle et al., 2004; Edwards et al., 2008).These impact examples were all significantly larger thanfireballs observed daily by the Desert Fireball Network(DFN) in Australia. Fireballs detected by the DFN haveenergies in the range of 10 to 10 J at atmosphericentry (Devillepoix et al., 2019). Meteorite-dropping fire-balls are at the upper energy range observed by theDFN.DFN is the world’s largest fireball camera network,located in the Australian outback and consisting of 52observatories, covering an area of 3 million km . It isaimed to detect fireballs, recover meteorites and to cal-culate their orbits (Devillepoix et al., 2019, 2018). Theobservatories are optimised to image objects having abrightness between 0 to -15 magnitudes which corre-sponds to sizes between 0.05 and 0.5 m (Devillepoixet al., 2019). In this work, we make a bulk seismic anal-ysis of the largest single data set of terrestrial fireballs obtained by the DFN in the period from 2014 to 2019,by systematically searching for seismic signals occurringin the time window and proximity of fireball trajectories.Unlike other studies that used data from images(Beech et al., 1995; Brown et al., 1994; Spurný et al.,2012), seismic stations (Brown et al., 2003; Devillepoixet al., 2020; Koten et al., 2019) and infrasound (El-Gabry et al., 2017) to calculate the orbits and energiesof meteors, this is the first study that uses informationabout the trajectory and timing of fireballs from a largedataset to back-trace any impact-related seismic activity.We investigate detection threshold of the DFN-observedfireballs in seismic data recorded by the Australian Na-tional Seismograph Network (ANSN). We also reporton the seismic properties of the fireballs caught by theseismic instruments. This information will be used forfuture instrument development in detecting fireballs inthe seismic domain. We used the DFN database containing trajectories of1410 fireball events that occurred above Australia overthe last 6 years. The DFN trajectory data provide abso-lute timing of fireball events, the start and end coordi-nates as well as the height above ground of the observedbright flight. A Python-based program was written tocalculate distances between the entire fireball trajectory(bright flight path) and all ANSN seismic stations. Theprogram was applied to all 1410 DFN fireballs. Thearrival times for the airwave are then calculated forboth the longest and the shortest direct distances, usinga speed of sound of 300 ±
60 m/s. We used this errormargin to account for local temperature and wind depen-dencies (Le Pichon et al., 2008). The large time windowalso considers unknown coupling with the ground andthe low signal strength.Seismic data were acquired from the ANSN, operatedby Geoscience Australia (GA), via public service domainIRIS (Incorporated Research Institutions for Seismol-ogy). The ANSN consists of a network of broadbandseismometers across Australia and its offshore territories.Figure 2 shows the locations of broadband seismometers(red triangles) and the coverage of DFN observatories(blue circles).The criteria that determined if a signal in time seriesdata can be confidently classified as a signal comingfrom a fireball event are:1. The amplitude of the signal must be similar or lowerthan previously confirmed seismic signals from fire-balls or bolides, accounting for uncertainties relatedto the event’s distance to a detector, yet above thebackground noise;2. The seismic signal must be within the calculated ar-rival times of the airwave (direct or ground-coupled tatistical analysis of fireballs: Seismic signature survey Figure 2.
Locations of GA seismometers (red triangles) andDFN camera observatories (blue circles). Some stations are closetogether and therefore symbols overlap.
Rayleigh wave; No P and S waves were identifiedin this survey);3. There must not be an earthquake activity in thedatabase (Geoscience Australia, 2019) at about thesame time;4. There must not be any clear anthropogenic-relatednoise (e.g., mine blasts, proximity to airport run-ways, etc). We note that DFN detects only night-time fireballs and at that time the anthropogenicnoise is expected to be minimal.The seismic time series data were obtained from thenearest seismic stations and checked for distinguishablesignals in the time window of the arrival of the airwaveand Rayleigh wave (Criteria 2). Time series data wereinterrogated for a time window starting 30 seconds priorto the start of a fireball event in the upper atmosphereand ending up to 28 minutes later. This is to accountfor the travel time of the airwave from the fireball toany seismic station within 400 km. The seismic datawas downloaded from the IRIS database. The Pythonframework ObsPy (Beyreuther et al., 2010; Krischeret al., 2015) was used to manipulate and analyse the timeseries data and the Python library Astropy (AstropyCollaboration et al., 2013, 2018) was used for makingcoordinate transformations. The time series data werefiltered using a Butterworth-Highpass filter at a defaultfrequency of 2 Hz. For most signals this filtering was themost satisfactory in cutting out ambient noise.In attempt to distinguish between meteor fragmenta-tion and the Mach cone passage, we used two approaches.We looked into the fireball orientation with respect to thelocation of the seismic station. If the shortest distance tothe seismic station is perpendicular to the bright flight trajectory and arrival time for the airwaves fits, signalsare classified as likely originating from the Mach cone. Ifthe shortest distance is not perpendicular to the brightflight trajectory, any seismic signals can be assumed tocome from a fragmentation along the trajectory. Consid-ering that the fragmentation has no preferred orientation,the events flagged as likely originated from the Machcone could have instead originated from the airburstcaused by fragmentation. However, we class them asMach cone events because previous literature reportedfragmentation to cause lesser air disturbance comparedto the Mach cone passage (Brown et al., 2003; Edwardset al., 2008). We also visually investigated DFN fireballimages to identify the distinct presence of fragmenta-tion. However, we were unable to unambiguously makesuch a distinction for all fireball events. This is prob-ably due to camera sensor saturation and because ofDFN cameras using the deBruin shutter sequence tomark absolute timing which interrupts visual light curverecording (Table 3).
Compared to larger impact events, it was expected thatthe DFN-observed fireballs could only cause occasionalweak seismic signals, predominantly coming from the at-mospheric disturbance, and only in favourable positionsand locations. Such an expectation was set by previousworks (Brown et al., 2003; Edwards et al., 2008).Table 1 shows the fireball events with suspected seis-mic signals including the start time of the bright flightobservation. Seismic signals were found for 25 fireballevents (Tables 1-3) out of 1410 surveyed, setting thedetectability at 1.8% when using the public seismic data.From here on, we will refer to specific events with theirallocated ID letter, rather than DFN event code name,as introduced in Table 1.Figure 3 shows the location of all DFN observatories(blue circles) and seismic stations of the ANSN (redtriangles) that identified these 25 events. It also showsthe trajectories of the bright flight of the fireballs forwhich seismic signals are suspected (yellow lines).Table 2 shows the coordinates of the beginning (lat b ,long b ) and the end (lat e , long e ) of the bright flight, thebeginning (h b ) and end (h e ) height, the trajectory slope,and the velocity (V), inferred mass (m) and fireballenergy (KE) at atmospheric entry. The slope is definedas the angle between the beginning of the bright flighttrajectory and local horizontal. The recorded fireballshad almost the entire range of possible impact angles(from 4 ◦ to 78 ◦ ) with a mean value ( ± σ ) of 38 ◦ ± ◦ .The mean h b was 86 ±
25 km and h e was 46 ±
18 km. Theimpact speed at the atmospheric entry was 25 ±
13 km/s.Meteoroids had a very large mass range, from 1 g up to180 kg estimated at atmospheric entry, correspondingto energies of 10 to 10 J. Neidhart et al.
Table 1
Fireball events with suspected seismic signals. Timeof fireball marks the start of the bright flight as observedby the DFN. Notation [A:Y] is to be used for easier crossreferencing between tables in this paper only.
Figure 3.
Locations of seismic stations in Australia of the ANSN(red triangles) which detected seismic signals from fireballs, DFNobservatories that observed fireballs that showed seismic signals(blue circles) and trajectories of the bright flight of fireballs forwhich suspected seismic signals have been detected (yellow lines).
Figure 4.
Time series data and spectrogram in vertical directionfor the only fireball event (DN160830_02) for which signals ofthe airwave and the Rayleigh wave can be identified separately.Signal was detected at the stations BBOO and high pass filterwas applied at 2 Hz. tatistical analysis of fireballs: Seismic signature survey Table 2
Fireball events with suspected seismic signals. Data includes the coordinates of the start (lat b , long b ) and end (lat e ,long e ) of the bright flight trajectory, initial velocity (V), inferred mass (m), and fireball energies (KE) at the top of theatmosphere and slopes (with respect to the horizon) as observed by the DFN. The uncertainties in the trajectory positionsare 0.1 km and the velocity uncertainties are 0.1 km/s. Masses are calculated using the dynamic method of Sansom et al.(2019) and are correct to an order of magnitude. Fireball energy is calculated as the transfer of kinetic energy on entry. ID lat b , long b lat e , long e h b h e V m KE Slope( ◦ , ◦ ) ( ◦ , ◦ ) (km) (km) (km/s) (g) (MJ) ( ◦ )A -27.4556, 135.1753 -27.5068, 134.3404 86.5 29.8 22.0 1200 286 34B -31.5395, 135.0993 -31.6170, 134.8425 75.5 42.9 21.4 10 2.29 51C -30.4119, 137.8785 -29.7315, 137.7925 69.6 37.3 15.3 900 105 23D -28.6623, 135.3836 -28.7460, 135.2587 75.1 41.7 18.0 80 13 65E -30.6361, 128.1459 -30.6801, 127.2522 99.2 72.2 38.9 2 1.51 17F -29.5592, 140.2430 -31.3590, 136.5810 98.6 66.8 38.3 10 7.33 4G -31.8641, 136.6595 -31.7352, 135.8630 111.0 90.1 69.9 1 2.44 15H -31.1390, 137.6323 -29.8901, 138.2426 85.6 31.4 17.2 1500 222 19I -30.0159, 137.9459 -30.1322, 138.6963 79.2 41.4 15.6 50 6.08 27J -28.5360, 134.7298 -28.4699, 134.5781 60.4 39.5 38.3 4700 3450 51K -28.2104, 135.5600 -27.4458, 135.3177 76.4 30.0 14.8 5500 602 28L -27.7673, 141.4574 -28.3419, 141.2917 91.9 54.9 34.9 5 3.05 29M -30.1016, 133.7398 -30.9163, 134.3092 88.2 58.3 14.4 30 3.11 16N -32.9211, 136.9160 -32.7568, 136.9180 57.3 45.0 18.3 400 67 34O -28.3672, 140.0076 -28.3798, 139.6592 97.2 70.7 31.5 60 29.80 37P -31.4627, 136.4002 -31.8320, 136.5779 82.7 42.0 13.7 5800 544 42Q -26.1650, 140.4517 -26.6136, 139.9639 90.4 29.8 29.3 900 408 41R -33.1445, 117.1267 -32.6246, 116.8404 82.4 25.6 16.4 5300 713 42S -32.9373, 138.8363 -32.7216, 138.9215 184.6 63.0 16.3 40 5.31 78T -27.8922, 137.0177 -27.2644, 136.6992 78.1 19.1 13.1 180000 15400 37U -28.0908, 136.2213 -27.8967, 136.3981 76.1 56.6 23.3 7 1.90 35V -28.0967, 136.4564 -28.0936, 136.5267 54.3 46.0 34.7 1900 1140 50W -31.1292, 136.2143 -31.2810, 136.2115 83.9 27.6 22.6 1500 383 73X -29.4472, 138.1771 -29.7606, 137.9361 94.6 54.5 35.0 5 3.06 43Y -33.3076, 119.5644 -33.4876, 119.4131 76.1 29.7 16.4 877 207.43 62 Neidhart et al.
Table 3
Fireball events with suspected seismic signal data, including the shortest station-to-trajectory distance (d min ), peakvalues for the seismic acceleration in vertical (BHZ), N-S (BHN) and E-W (BHE) seismic axes, estimated duration of theseismic signal (t), and peak frequency ( ν ) after applying 2 Hz high pass filter. Based on the arrival times, the seismic sourcecan be a direct airwave (A) or a ground-coupled Rayleigh wave (R). The last column shows whether the optical image of thefireball displayed clear evidence of fragmentation processes. *Note that NWAO station is non-aligned to cardinals. ID Station d min
BHZ × − BHN × − BHE × − t ν Source Fragmentation(km) (mm/s ) (mm/s ) (mm/s ) (s) (Hz)A OOD 106.8 1.38 0.62 0.67 16 3 A and/or R yesB BBOO 180.7 3.91 6.54 1.78 12 4-10 A noC LCRK 74.2 9.79 5.71 6.56 7 3 A or R noD OOD 121.3 3.47 1.03 1.17 3 3 A yesE FORT 93.0 5.04 0.67 0.17 7 3 A and/or R yesF LCRK 78.9 3.21 2.22 2.88 20 3 A or R -G BBOO 150.3 1.01 0.26 0.66 9 3-5 A or R noH LCRK 53.4 13.0 1.38 6.34 25 3-6 A or R yesI LCRK 69.4 9.22 3.52 4.64 9 3-5 A or R noJ OOD 138.1 3.47 1.37 1.84 7 3 A yesK OOD 54.4 14.5 3.08 2.38 37 3-5 A or R noL INKA 100.3 17.4 4.26 6.56 55 3 A or R -M MULG 77.4 17.5 15.7 7.14 16 3-4 A or R yesN BBOO 92.7 0.71 0.36 0.35 24 3-5 A -O INKA 140.3 6.29 2.96 2.00 18 3-4 A yesP BBOO 126.5 3.56 1.82 0.92 17 3-5 A and/or R yesQ INKA 150.3 7.91 3.43 3.29 8 3-5 A yesR MUN 96.8 0.94 1.16 0.76 10 3-5 A -S HTT 101.1 3.08 1.69 3.22 7 3-4 A or R -T OOD 117.6 5.68 3.41 1.98 11 3-10 A -U OOD 90.9 3.31 2.53 1.64 8 3-4 A or R -V OOD 99.3 2.13 2.19 2.25 6 3-4 A yesW BBOO 172.7 0.62 0.82 1.03 12 3-4 A yesX LCRK 97.7 0.51 0.59 0.54 43 3-5 A or R yesY NWAO* 214.5 1.47 - - 20 7-10 A no tatistical analysis of fireballs: Seismic signature survey min ),the peak values for the acceleration in vertical (BHZ),N-S (BHN) and E-W (BHE) components seen in thetime series data, the duration of the signal (t), the peakfrequency ( ν ) and estimates for the seismic source. Theseismic signals for all 25 fireballs are between 3 s and55 s long and the peak values of the seismic frequenciesare up to 10 Hz with an average at 3.8 ± ±
40 km, ranging from 53 km to 215km, although the surveyed area reached the maximumof 325 km distance. No surveyed fireballs were detectedby more than one seismic station. This is expected giventhe sparse distribution of ANSN stations and is roughlyin agreement with previous works (Brown et al., 2003,2004).Figure 5 shows the time series data for 25 fireballevents [A:Y] for which seismic signals were detected. Itcan be seen that for 18 out of 25 events, the highestpeaks are in the vertical direction. We examined anycorrelations between the direction of the highest peak inamplitude seen in the time series data and the positionof the seismic station relative to the trajectory of thefireball and if the fireball approaches the seismic stationor not. However, we did not find any other azimuth ondirectionality. On average, the amplitude for the highestpeaks for seismic signals in the vertical direction was5.5 × − mm/s , while it was 2.7 × − mm/s in N-Sand 2.4 × − mm/s in E-W directions. This suggesteda slight preference in vertical direction agreeing withthe assumption that the seismic excitation was from theatmosphere.Figure 6 shows the highest peak in vertical directionas a function of the shortest distance between the tra-jectory and the seismic station for all events for whichseismic signals are suspected. The colours of the markersrepresent the slope of the fireballs. It can be seen thatfireballs that occur very close to the seismic station havehigher peak amplitudes in vertical direction than fireballsfurther away. There is also additional observational biasthat could be attributed to favourable fireball orienta-tion to create Mach cone disturbance that is directed ata seismic station. The Mach cone-related fireball detec-tions are more likely to originate from shallower (lower)impact angles that assure longer trajectories in the at-mosphere than in the case of suspected fragmentationas a seismic source. From the 1410 DFN fireball events surveyed, we identifyseismic signals in time series data that correspond to 25of these events. This is 1.8%. Figure 6 shows there is a
Figure 5.
Highest peaks in time series data in vertical, North-South and East-West direction for 25 seismic signals that mightoriginate from the Mach cone of fireball events (A-M) (upper) andfrom fragmentation (N-Y) (lower). 18 signals show the highestpeak in vertical direction.
Figure 6.
Highest amplitude in vertical for all 25 fireballs forwhich seismic signals are suspected as a function of the shortestdistance between bright flight trajectory and seismic station. Thecolours of the markers show the slope of these fireballs. The peakamplitude is decreasing with distance to the seismic station.
Neidhart et al. rough correlation between peak amplitude and distanceto a seismic station. Beyond 215 km we do not detectany unambiguous seismic signals, and the furthest eventsare all steep-sloped. It is therefore reasonable to placea threshold at 215 km as an approximate limit for theseismic detection of fireballs. Given that, the numberof DFN fireballs within this range is reduced to 1101,increasing the detection success to 2.3%. DFN observa-tories are approximately 150 km apart. There were 1236of fireball trajectories within 215 km of a DFN obser-vatory. Should the DFN camera network be equippedwith seismic instruments (of comparable sensitivity) ateach observatory site, 86% of observed fireballs would bewithin the 215 km distance threshold for detection in theseismic domain. The mean distance to a seismic stationof detected fireballs using ANSN was 112 km (Fig. 6),which corresponds to about 50% of all surveyed fireballsif each observatory site had a seismic station equipped. Itwould be possible to detect fireballs at multiple stations,with an average of four stations per fireball.The survey showed that some seismic stations are moresensitive to fireball events than others. The highest num-ber of signal detections was at the station Oodnadatta(OOD) which detected 7 suspected fireball events fol-lowed by Buckleboo (BBOO) and Leigh Creek (LCRK),where each detected 5 events, and Innamincka (INKA)with 3 events. There are five seismic stations (Forrest(FORT), Mundaring (MUN), Hallett (HTT), Mulgath-ing (MULG), Narrogin (NWAO)) that only detected oneevent. This could be due to the individual instrumentquality or background noise levels which are influencedby the positioning setup and geographic location of thesensor. Previous studies by Revelle et al. (2004) have alsopointed this out. Another sensitivity to detection mightbe directionality between seismic stations and brightflight trajectory. Seismic stations that are perpendicularcan detect the signal from the Mach cone which has ahigher amplitude and is therefore easier to recognize. Acombination of these factors, like the presence of noise,distance to the station, the directionality from the tra-jectory to the seismic stations, weather conditions, soilproperties and also the characteristics of the impactor,are among reasons we did not detect more than 2.3%events within the 215 km threshold.As well as identifying the 25 fireball events in seismictime series data, we also investigated five of the largestevents ever seen by the DFN. Unfortunately none passthe selection criteria. To date, there are two eventsdetected by the DFN (DN150102_01, DN170630_01)that have also been recognized by the US GovernmentSensors (USG) and described in detail by Devillepoixet al. (2019). The closest stations to these two eventswhere data are available were 120 and 182 km away.These stations show noisy signals or a signal only in onecomponent.We also looked for seismic signals from fireballs that had dropped a meteorite (Murrili, Sansom et al. (2020);Dingle Dell, Devillepoix et al. (2018); DN160822_03,Shober et al. (2019)) that were recovered from the field.The closest stations to these events were 150; 93 and169; and 191 km respectively and show noisy seismicdata and no signals.
Fireball events occur on a daily basis, yet are rarely re-ported as seismic events because their energy (at the topof the atmosphere) is often not sufficient to cause quakesthat are detectable by seismic stations. Unlike otherstudies who used data from images, seismic stations andinfrasound to calculate the orbit and energies of meteors,this study uses information about the trajectory andtiming of fireballs observed by the DFN to search forseismic signals.We report possible detections of 25 seismic signaturesoriginating from 1410 surveyed fireballs observed by theDFN over a 6-year period. This is made by calculat-ing the distance between the bright flight trajectory ofthe fireball to Australian National Seismograph Network(ANSN) seismic stations. We searched for significant seis-mic signals recorded that fit our selection criteria. Theobserved signals cannot be explained to be of any othergeologic or anthropogenic origin. Signals are seconds-long in duration and have peak amplitude ranges in thefollowing components:• Vertical: 5 × − mm/s – 2 × − mm/s • N-S: 3 × − mm/s – 2 × − mm/s • E-W: 4 × − mm/s – 7 × − mm/s The total of 18 out of 25 signals showed the highestpeak in vertical component. The signals showed the peakfrequency in the range up to 10 Hz. Calculations of ar-rival times suggests signals are due to direct airwaves orground-coupled Rayleigh waves. The fireball directional-ity suggest that about half of the observed signals couldhave been caused by the Mach cone and the other halforiginated from fragmentation of the impactor.We propose an upper threshold for seismic detectabiltyof fireballs to be approximately 215 km. If a seismome-ter (of equal sensitivity) was installed alongside thesesystems, it may have been possible to record 50% of allDFN fireballs.