Combined use of repeated active shots and ambient noise to detect temporal changes in seismic velocity: application to Sakurajima volcano, Japan
© The Author(s) 2017
Received: 4 September 2016
Accepted: 31 January 2017
Published: 13 March 2017
KeywordsCoda-wave interferometry Seismic interferometry Seismic velocity change Active seismic experiment Ambient noise Sakurajima volcano
Temporal changes in seismic velocity associated with large earthquakes or volcanic activities have been detected from analyses of coda-wave interferometry or seismic interferometry. These interferometric techniques are now widely used to monitor temporal changes in structure. Coda-wave interferometry is a technique to detect slight changes in medium properties by measuring phase changes or waveform changes in coda waves from repetitive similar sources (e.g., Poupinet et al. 1984; Snieder et al. 2002; Snieder 2006). Poupinet et al. (1984) first applied coda-wave interferometry to an earthquake doublet consisting of two micro-earthquakes that occurred 14 months before and 7 months after the M5.9 Coyote Lake, California, earthquake, and detected seismic velocity decrease of 0.2%. Nishimura et al. (2000, 2005) analyzed records of repeated active seismic experiments conducted at Iwate volcano, Japan, from 1998 to 2003, and detected seismic velocity decrease of about 1%. These two studies interpreted that these velocity decreases were caused by static stress change in shallow structures close to M5.9 or M6.1 earthquakes and/or volcanic pressure sources. Cannata (2012) applied coda-wave interferometry to repeating volcano-tectonic earthquakes that occurred with about 3-h interval at Mt. Etna and obtained detailed temporal seismic velocity changes in the volcanic edifice. However, such a high time resolution is rare in general, because the time intervals of repeating earthquakes or repeated active seismic experiments are usually long in time.
On the other hand, seismic interferometry, which is a technique to estimate relative seismic velocity changes using cross-correlation functions of ambient seismic noise (e.g., Shapiro and Campillo 2004; Curtis et al. 2006), enables us to continuously monitor seismic velocity changes. Brenguier et al. (2008) determined seismic velocity changes at Piton de la Fournaise volcano every 10 days. They interpreted the observed seismic velocity decrease of about 0.1% preceding eruptions at the volcano by dilatation of the edifice due to pressure increase in the magma. Takagi et al. (2012) detected a coseismic decrease in Rayleigh wave velocity of 0.1–0.5% related to the occurrence of the 2008 Iwate–Miyagi Nairiku earthquake, NE Japan (M7.2). Since ambient noises are available everywhere, seismic interferometry has been applied at many fields for continuous monitoring (e.g., Sens-Schönfelder and Wegler 2011). However, ambient noise source distribution may affect the quality of retrieved Green’s functions and accordingly relative seismic velocity changes. For example, Stehly et al. (2006) demonstrated that noise sources show a clear seasonal variation; the noise sources in 0.05–0.1 Hz are located in the northern Atlantic and the Pacific during the winter in the northern hemisphere, while in the Indian Ocean and in the Southern Pacific during the summer in the northern hemisphere.
Recently, Budi-Santoso and Lesage (2016) applied these two techniques simultaneously. They detected seismic velocity decrease of up to 1.5% associated with the 2010 eruption at Merapi volcano, Indonesia. Temporal changes in seismic velocity estimated from seismic interferometry analysis displayed non-synchronized changes with those from coda-wave interferometry analysis. Such non-synchronized changes may be caused by the difference of wave fields because the coda-wave interferometry used volcano-tectonic earthquakes occurring at depths of about 0.5–2.3 km and not at the ground surface. In addition, the errors on vertical locations of volcano-tectonic earthquakes were as large as about a few hundred meters. Cross-correlation coefficients of the similar seismograms were higher than 0.75. However, that threshold is a little bit lower.
Sakurajima is one of the most active volcanoes in Japan, which is located at the southern rim of Aira caldera in southern Kyushu, Japan. The volcano is well known to exhibit vulcanian eruptions in these decades, and the number of explosions mainly from Showa crater reached 835 in 2012, 883 in 2013, and 450 in 2014 (Japan Meteorological Agency 2015). The volcanic activities have been monitored by various observations: seismic and geodetic measurements as well as volcanic gas observations (e.g., Iguchi et al. 2013). Tsutsui et al. (2016) further repeated active seismic experiments once a year from 2008 and detected temporal changes in reflection coefficients that are inferred to be caused by density or seismic velocity changes in the sill at a depth of around 4.9 km beneath Sakurajima volcano. However, there is no study so far that applies coda-wave interferometry to the data of the repeated active seismic experiments at Sakurajima. There are only a few researches that continuously monitor temporal structure changes at Sakurajima; Maeda et al. (2015) reported temporal changes in the subsurface structure using waveforms of an active source called accurately controlled routinely operated signal system (ACROSS) during an active eruptive period from September 2012 to July 2014.
In this study, we apply coda-wave interferometry to the repeated active seismic experiments conducted at Sakurajima once a year from 2011 to 2014. Also, we apply seismic interferometry to ambient seismic noise at Sakurajima. Since active seismic sources are located near the ground surface, the seismic waves we analyze in these two kinds of methods are almost same and sample same regions. Most of shot points at each shot site are located within 10 m, and we choose records of which the cross-correlation coefficient is higher than 0.9 between the two shots. We compare seismic velocity changes obtained from these two methods and discuss a relation between the seismic velocity changes and the volcanic activity at Sakurajima.
Seismic interferometry is applied to the ambient seismic noise data between 2012 and 2014. Ambient noises are first filtered in 1–2, 2–4, and 4–8 Hz bands. For each frequency band, we choose 10-min-long data having amplitude of smaller than 10 times of the average RMS amplitude of ambient noise recorded in 2012. We apply 1-bit normalization and spectral whitening to reduce the effect of earthquake signals (e.g., Bensen et al. 2007). Finally, cross-correlation functions (CCFs) are calculated every 10 min, and daily CCFs (hereafter called DCCFs) are obtained by stacking the 10-min CCFs every day for 15 pairs from six stations. The obtained DCCFs are considered to be Green’s functions recorded at one of a station pair from a virtual source at the other station of the pair.
Estimation method of seismic velocity changes
To estimate seismic velocity changes, we use Moving Window Cross-Spectral (MWCS) technique (Poupinet et al. 1984). In this method, delay times, dt, between two similar seismograms are calculated at respective lapse times using a relation, \( \phi (f) = 2\pi fdt \), where \( \phi \) is the phase of cross-spectrum and f is the frequency, for consecutively overlapping short time windows. We set the length of the short time windows to be 2.56 s for 1–2 and 2–4 Hz, 1.28 s for 4–8 Hz, 0.64 s for 8–16 Hz, and 0.32 s for 16–32 Hz, respectively, which are overlapped by 75% with the adjacent time windows. Assuming a spatially homogeneous velocity change, we derive a relation of \( dt/t = - dv/v \), where v is the seismic velocity, and t is the lapse time. We estimate relative velocity changes dv/v from linear regression of the delay time with respect to the lapse time for the data whose cross-correlation coefficients are higher than 0.7 in the short time windows. Coherences in the short time windows and an example of estimating dt/t are shown in Fig. 2. Coherences become lower from around 10 s in lapse times. From this reason, records of the repeated active seismic experiments in lapse times between 0 s and 10 s are used in coda-wave interferometry analysis. In seismic interferometry analysis, DCCFs in lag times between −10 and +10 s are used. Each DCCF is cross-correlated with a reference CCF, which is calculated by stacking DCCFs over 2012 and 2013.
Mean values of relative velocity changes from the coda-wave interferometry analysis of the vertical-component records
Period of time
+0.016 ± 0.089% (7)
+0.038 ± 0.059% (20)
+0.044 ± 0.042% (17)
+0.130 ± 0.137% (4)
+0.049 ± 0.054% (4)
+0.046 ± 0.034% (2)
−0.064 ± 0.089% (4)
−0.030 ± 0.100% (7)
−0.012 ± 0.089% (4)
+0.225 ± 0.152% (4)
+0.144 ± 0.096% (3)
+0.002 ± 0.121% (13)
−0.004 ± 0.049% (15)
+0.018 ± 0.055% (8)
−0.008 ± 0.115% (9)
+0.027 ± 0.052% (12)
+0.015 ± 0.048% (15)
We only use records of which the cross-correlation coefficient is higher than 0.9. Hence, the obtained results are reliable. However, the dynamite charge of 20 kg for all the shots except one at Sakurajima restricts the estimation of the velocity changes to the northeastern flank of the volcano. This limitation of wave paths makes it difficult to compare the results obtained from coda-wave interferometry and those from seismic interferometry for almost identical wave paths (only a few exceptions are shown in Additional file 1). For this reason, we average the values of relative velocity changes for all shot–station pairs when we compare the results from coda-wave interferometry with those from seismic interferometry.
We compare the results obtained from coda-wave interferometry with those from seismic interferometry. Mean values of the relative velocity changes from coda-wave interferometry are calculated by averaging the results from the vertical component for all shot–station pairs. Results of seismic interferometry are also averaged for all 15 station pairs. In Fig. 6, the mean relative velocity changes shown by red squares are overwritten by assuming that the results for the 2012 experiment are the same as those on the same day from seismic interferometry. Error bars show ± one standard deviation. The mean relative velocity changes distribute within the variability among station pairs in seismic interferometry between 2012 and 2014 in 2–4 and 4–8 Hz band. Hence, the results from these two techniques are consistent with each other. This is also supported by comparing the results of coda-wave interferometry and those of seismic interferometry for only a few almost identical paths (Additional file 1).
We compared the velocity changes of 2011, 2013, 2014 to that of 2012 at first. However, other choices of a reference year and combination of two shot years are possible. Since there is no result from seismic interferometry in 2011, we cannot use 2011 as the reference year due to the unavailability of data. In Fig. 6a, we show the differences by changing the reference year. Pairs of shot years are fixed to be 2011–2012, 2012–2013, and 2012–2014. Different colors indicate the differences of the reference year (2012: red, 2013: blue, 2014: green). Blue line slightly shifts to the above from the other two. However, the values of mean relative velocity changes are still within pair variations in seismic interferometry between 2012 and 2014. The results of all three patterns are almost the same in 4–8 Hz band. In Fig. 6b, we show the results of various combinations of two shot years. Calculated mean relative velocity changes from 2013 to 2014 are shown by green, those from 2011 to 2014 by blue, and those from 2011 to 2013 by cyan. Plots by red indicate the velocity changes in each year to that of 2012. The differences between them are quite small both in 2–4 and in 4–8 Hz bands, at a maximum about 0.05% in 2–4 Hz band for the case of 2011–2013. Those comparisons indicate that we can obtain reliable results from coda-wave interferometry.
We have compared the mean values of velocity changes. To see the spatial variations of velocity changes, we need to consider sensitivity kernels (e.g., Pacheco and Snieder 2005). The kernels have two peaks at the position of the source and station (see Additional file 2). Therefore, measured velocity changes are expected to be sensitive to local changes at the source and station. Coda waves that we use are probably the mixture of body waves and surface waves. One advantage of our study is that both of shot points and stations are located at near surface. Therefore, the sensitivity kernels have large sensitivities near the surface regardless of the predominance of either body waves or surface waves. Almost annual changes in seismic velocity are detected commonly for many shot–station pairs and station pairs. This suggests that the almost annual changes in seismic velocity take place at shallow depths widely over Sakurajima. On the other hand, it is true that the spatial variation of the estimated relative velocity change exists, especially in the shorter-period component. So far, causes of such spatial variation are not clear yet, though it may reflect structural changes in the vicinity of shot points or stations.
We estimate strain sensitivities of seismic velocity changes using the areal strain values to be about 2.0 × 103/strain at 1–2 Hz. This sensitivity ranges within previously reported values that are compiled in Nishimura et al. (2005). Such consistency strongly suggests that observed annual changes in seismic velocity are related to strain changes in volcanic edifice. At Sakurajima, spherical pressure sources that explain observed ground deformation between November 2009 and April 2010 well are estimated at a depth of about 12 km beneath the Aira caldera (north of Sakurajima) and at a depth of about 5 km beneath Kita-dake (north part of Sakurajima) (e.g., Iguchi et al. 2013). However, there is no study which explains geodetic data in other periods of time as far as we know. Therefore, the reason why geodetic data show the annual-like change is not clear yet.
We examined seismic velocity changes at Sakurajima from 2011 to 2014 by combining coda-wave interferometry analysis of the repeated active seismic experiments and seismic interferometry analysis of ambient seismic noise. The sources of active seismic experiments are located near the ground surface, which enabled us to directly compare seismic velocity changes from these two techniques. We compare the results obtained from two techniques for all reference years and combinations of two shot years. In all cases, the results from these two techniques are consistent with each other. The values of mean relative velocity changes in coda-wave interferometry are within the variability among station pairs in seismic interferometry between 2012 and 2014 in 2–4 and 4–8 Hz band. Almost annual change in relative velocity changes is clearly found in both of the estimated seismic velocity changes and geodetic signals from GNSS receivers, which is a clear example that seismic velocity is well correlated with geodetic signals at a volcano. The combined use of coda-wave interferometry and seismic interferometry is shown to be effective to obtain reliable and continuous measurements of seismic velocity changes. This technique will be useful as one of monitoring tools of volcanoes.
TH (corresponding author) carried out the analyses and drafted the manuscript. All authors read and approved the final manuscript.
We used seismograms obtained from active seismic experiments conducted by Disaster Prevention Research Institute, Kyoto University, other eight Japanese universities, and Japan Meteorological Agency (JMA) at Sakurajima. Especially, we are very grateful to Profs. Tomoki Tsutsui and Masato Iguchi for their leadership to conduct the active seismic experiments for as long as 6 years from 2008 to 2014. Our coda-wave interferometry analysis would not be possible without their great efforts. We are also grateful to JMA for providing us with continuous seismic data recorded at 6 stations. We thank Geospatial Information Authority of Japan for making available GNSS data and Digital Elevation Model data with a mesh size of 10 m. We appreciate many constructive comments from an associate editor, Prof. Hiroshi Takenaka, and two anonymous reviewers. This study was supported by Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan, under its Earthquake and Volcano Hazards Observation and Research Program.
The authors declare that they have no competing interests.
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.
- Bensen GD, Ritzwoller MH, Barmin MP, Levshin AL, Lin F, Moschetti MP, Shapiro NM, Yang Y (2007) Processing seismic ambient noise data to obtain reliable broad-band surface wave dispersion measurements. Geophys J Int 169:1239–1260View ArticleGoogle Scholar
- Brenguier F, Shapiro NM, Campillo M, Ferrazzini V, Duputel Z, Coutant O, Nercessian A (2008) Towards forecasting volcanic eruptions using seismic noise. Nat Geosci 1:126–130View ArticleGoogle Scholar
- Budi-Santoso A, Lesage P (2016) Velocity variations associated with the large 2010 eruption of Merapi volcano, Java, retrieved from seismic multiplets and ambient noise cross-correlation. Geophys J Int 206:221–240View ArticleGoogle Scholar
- Cannata A (2012) Crustal changes at Mt. Etna volcano accompanying to 2002–2003 eruption as inferred from a repeating earthquake analysis. Geophys Res Lett 39:L18311. doi:10.1029/2012GL0531853 View ArticleGoogle Scholar
- Curtis A, Gerstoft P, Sato H, Snieder R, Wapenaar K (2006) Seismic interferometry—turning noise into signal. Lead Edge 25:1082–1092. doi:10.1016/j.jvolgeores.2006.04.003 View ArticleGoogle Scholar
- Froment B, Campillo M, Roux P, Gouedard P, Verdel A, Weaver RL (2010) Estimation of the effect of nonisotropically distributed energy on the apparent arrival time in correlations. Geophysics 75:SA85–SA93. doi:10.1190/1.3483102 View ArticleGoogle Scholar
- Iguchi M, Tameguri T, Ohta Y, Ueki S, Nakao S (2013) Characteristics of volcanic activity at Sakurajima volcano’s Showa crater during the period 2006 to 2011. Bull Volcanol Soc Jpn 58:115–135Google Scholar
- Japan Meteorological Agency (2015) (1) Sakurajima. Documents for the 132th meeting of the coordinating committee for prediction of volcanic eruptions (in Japanese). http://www.data.jma.go.jp/svd/vois/data/tokyo/STOCK/kaisetsu/CCPVE/shiryo/129/132_no02.pdf. Accessed 31 Mar 2016
- Maeda Y, Yamaoka K, Miyamachi H, Watanabe T, Kunitomo T, Ikuta R, Yakiwara H, Iguchi M (2015) A subsurface structure change associated with the eruptive activity at Sakurajima Volcano, Japan, inferred from an accurately controlled source. Geophys Res Lett 42:5179–5186View ArticleGoogle Scholar
- Miyamachi H, Tomari C, Yakiwara H, Iguchi M, Tameguri T, Yamamoto K, Ohkura T, Ando T, Onishi K, Shimizu H, Yamashita Y, Nakamichi H, Yamawaki T, Oikawa J, Ueki S, Koeda T, Masuda Y, Katou K, Hatakeyama K, Kobayashi T (2013) Shallow velocity structure beneath the Aira caldera and Sakurajima Volcano as inferred from refraction analysis of the seismic experiment in 2008. Bull Volcanol Soc Jpn 58:227–237Google Scholar
- Nishimura T, Uchida N, Sato H, Ohtake M, Tanaka S, Hamaguchi H (2000) Temporal changes of the crustal structure associated with the M6.1 earthquake on September 3, 1998, and the volcanic activity of Mount Iwate, Japan. Geophys Res Lett 27:269–272View ArticleGoogle Scholar
- Nishimura T, Tanaka S, Yamawaki T, Yamamoto H, Sano T, Sato M, Nakahara H, Uchida N, Hori S, Sato H (2005) Temporal changes in seismic velocity of the crust around Iwate volcano, Japan, as inferred from analyses of repeated active seismic experiment data from 1998 to 2003. Earth Planets Space 57(6):491–505. doi:10.1186/BF03352583 View ArticleGoogle Scholar
- Pacheco C, Snieder R (2005) Time-lapse travel time change of multiply scattered acoustic waves. J Acoust Soc Am 118(3):1300–1310View ArticleGoogle Scholar
- Poupinet G, Ellsworth WL, Frechet J (1984) Monitoring velocity variations in the crust using earthquake doublets: an application to the Calaveras Fault, California. J Geophys Res 89:5719–5731View ArticleGoogle Scholar
- Sens-Schönfelder C, Wegler U (2011) Passive image interferometry for monitoring crustal changes with ambient seismic noise. CR Geosci 343:639–651View ArticleGoogle Scholar
- Shapiro N, Campillo M (2004) Emergence of broadband Rayleigh waves from correlations of the ambient seismic noise. Geophys Res Lett 31(7):L07614. doi:10.1029/2004gl019491 View ArticleGoogle Scholar
- Snieder R (2006) The theory of coda wave interferometry. Pure appl Geophys 163:455–473View ArticleGoogle Scholar
- Snieder R, Grêt A, Douma H, Scales J (2002) Coda wave interferometry for estimating nonlinear behavior in seismic velocity. Science 295(5563):2253–2255View ArticleGoogle Scholar
- Stehly L, Campillo M, Shapiro NM (2006) A study of the seismic noise from its long-range correlation properties. J Geophys Res 111:B10306. doi:10.1029/2005jb004237 View ArticleGoogle Scholar
- Takagi R, Okada T, Nakahara H, Umino N, Hasegawa A (2012) Coseismic velocity change in and around the focal region of the 2008 Iwate–Miyagi Nairiku earthquake. J Geophys Res 117:B06315. doi:10.1029/2012jb009252 View ArticleGoogle Scholar
- Tsai VC (2011) A model for seasonal changes in GPS positions and seismic wave speeds due to thermoelastic and hydrologic variations. J Geophys Res 116:B04404. doi:10.1029/2010JB008156 Google Scholar
- Tsutsui T, Iguchi M, Nakamichi H, Tameguri T, Nakamichi H (2016) Structural evolution beneath Sakurajima volcano, Japan, revealed through rounds of controlled seismic experiments. J Volcanol Geotherm Res 315:1–14. doi:10.1016/j.jvolgeores.2016.02.008 View ArticleGoogle Scholar