Reverberations, coda waves and ambient noise: Correlations at the global scale and retrieval of the deep phases

https://doi.org/10.1016/j.epsl.2014.01.047Get rights and content

Highlights

  • Body-waves retrieval from cross-correlation of continuous records.

  • Global scale interferometry in the 5 s to 100 s period band.

  • Impact of reverberated waves from large earthquakes.

Abstract

Cross-correlation of continuous broadband records allows the retrieval of body waves at teleseismic distances. These continuous records mainly contain low-amplitude background noise that comes from ocean–crust interactions, although there are also many transient events of different magnitudes and their coda associated with reverberation and/or scattering. We present an analysis at the global scale of these different contributions in the context of body-wave retrieval using the cross-correlation technique. Specifically, we compare the correlation of long codas after strong earthquakes with those of the quietest days. In the long period range (25–100 s), several phases that propagate in the deep Earth are observed in the correlations of the signals recorded after earthquakes, with some of these phases showing non-physical polarization. At the same time, the global section of correlations shows a series of spurious branches. These features are reproduced with synthetic correlations. A stack of the quietest days of the year shows that body waves are still present, with relative amplitudes that are closer to those expected for the actual Earth response. When considering shorter periods (5–10 s), the reconstruction of the deep phases is not affected by the earthquake coda, due to the dominance of scattering over reverberation.

Introduction

From the regional to the global scale, ambient seismic noise primarily refers to the wavefield that is continuously produced by the interactions of the fluid envelopes, as mainly through ocean waves, and the solid Earth. The source mechanisms of these interactions are frequency dependent. Short-period noise (4 s to 30 s) is dominated by the two microseism peaks (e.g., Longuet-Higgins, 1950, Ardhuin et al., 2011). At longer periods (above 30 s), other mechanisms take place, which are also known as Earth hum (Kedar and Webb, 2005), such as the proposed shear-wave generation by infragravity waves (Fukao et al., 2010). Here the term “noise” is defined through its difference from the earthquake records. The duration of an earthquake record is defined with respect to a particular signal-to-noise ratio (SNR) threshold, and it varies with frequency for a given event magnitude. Furthermore, depending on the frequency, the scattering strength governs the ratio between the randomly scattered waves and the ballistic waves that reverberate between the main boundaries (i.e., the Earth surface, the core–mantle boundary).

It has been demonstrated that the elastic response between two stations can be evaluated by correlation of the records of scattered waves (Campillo and Paul, 2003) or long ambient noise records (Shapiro and Campillo, 2004). As expected from the theoretical Green's function between two points at the free surface, the correlations of continuous records are dominated by surface waves. The application of this approach has led to numerous examples of surface-wave imaging (e.g. Shapiro et al., 2005, Sabra et al., 2005a, Ritzwoller et al., 2011). The extension of the approach to body waves is indeed appealing, although the level of the remaining random fluctuations in the correlations makes the identification and exploitation of weak signals difficult. Furthermore, the sources of ambient noise are likely located at the surface, which results in a dominance of surface waves in the noise records. However, teleseismic body-waves have been observed in noise records (e.g. Vinnik, 1973, Gerstoft et al., 2008, Landès et al., 2010). The search for body waves in the correlations has been successful in the last few years, which started with the crustal phases (Zhan et al., 2010, Ruigrok et al., 2011, Poli et al., 2012a). Then, deep vertical reflections were detected from the mantle transition zone (Poli et al., 2012b) and from the core (Lin et al., 2013), with data from regional arrays. The complete teleseismic section was reconstructed by cross-correlation using a worldwide combination of arrays at short to long periods (5 s to 100 s; Boué et al., 2013) and at long to very long periods (30 s to 300 s; Nishida, 2013). These last studies demonstrated the feasibility of ambient noise body-wave imaging. Lin and Tsai (2013) also discussed core-phase retrievals using antipodal station pairs.

Different processing has been used in all of these studies, and especially regarding the removal of transient signals. Nishida (2013) applied the most rigorous processing, using the Global Centroid Moment Tensor catalog (Ekström et al., 2012) to systematically remove long time windows corresponding to earthquakes and the following few days, the number of which depended on the event magnitude (Nishida and Kobayashi, 1999). Lin et al. (2013) and Boué et al. (2013) used less restrictive criteria. At the global scale, both Nishida (2013) and Boué et al. (2013) observed mantle body waves but obtained different results for the amplitudes of the core phases, which are weaker, and were more realistic in the correlation computed by Nishida (2013). Note that large-amplitude core phases were also reported by Lin et al. (2013).

Boué et al. (2013) questioned the relevance of these high-amplitude phases, and suggested that they show non-physical features. By non-physical, we mean here that these features do not appear in the natural Green's function. For example, the phase in the correlation corresponding to ScS is observed at short distances with strong amplitudes for the vertical component, which leads to an obvious problem of polarization. The presence of spurious arrivals in the correlation section challenges the applicability of noise imaging to body-wave problems in the deep Earth. We address this problem here, by analyzing the conditions under which reliable information can be extracted from noise correlations.

On the other hand, Lin et al. (2013) observed a strong correlation between the phases that reach the deepest parts of the Earth (ScS, PPdf) and the seismicity. They suggested that earthquakes mainly excite these body waves. Finally the observations of Lin et al. (2013) and Boué et al. (2013) included spurious phases that are not in the Earth response, or at least, have different relative amplitudes. The problem of spurious arrivals due to multiples was discussed on a smaller scale by Snieder et al. (2006). Concerning wave propagation at the global scale, Ruigrok et al. (2008) discussed the imperfect reconstruction of the Green's function from surface source records, which reveals the presence of spurious arrivals (ghost events). They derived an elastodynamic relation from the representation theorem showing that knowledge of the responses of the medium with and without the effects of the free surface is required to retrieve the exact Green's function. They verified this theoretical statement numerically with acoustic simulations.

By investigating the temporal evolution of the reconstructed Green's function after large seismic events, it is shown in the present study that the processing used can explain these observations at long periods. The structure of this report is the following. First, the dataset used and the processing applied are shown. Then we compare the quality of the reconstruction of some of the phases with the seismicity and the microseism excitation over a whole year. We present a synthetic example of the reconstruction of the partial Green's functions using a simulated long time reverberated coda wavefield to explain the characteristics of spurious arrivals. Finally, the study focuses on the particular propagation geometry between Finland and Japan.

Section snippets

Data and processing

In this study, one year was selected (2008) for the vertical-component records from a set of 420 stations distributed worldwide (Fig. 1). The BH channels are used after removal of the instrumental response, and decimation to a 5-Hz sampling frequency. Note that some of these stations are not available during the whole of the year period. All of the networks involved are detailed in Appendix A. The continuous records were processed similarly to Boué et al. (2013), which includes spectral

Contributions to correlations

We construct daily global sections that are used to evaluate the temporal evolution of the contributions of the daily correlations to the reconstructed Green's function. The daily contributions are quantified by computing the coherence between the daily and yearly reconstructions in specific time–distance domains. We selected six different time–space domains that correspond to different phases (Fig. 2b, colored boxes), and processed the image correlation for the 366 days of the year. This

Long-period processing

Using the long-period band, the coherence is presented in Fig. 3d, where we define high-coherence days (HCDs) by selecting the days with high seismic activity that are also associated with high coherence for each phase. In practice, we selected days with a coherence >0.2 and with a corresponding local maximum in the cumulative seismic moment function. For the mantle P-waves and S-waves, this results in selecting a group of days that correspond to the high-amplitude peaks in February and March.

A specific geometry: the FNET–LAPNET dataset

The FNET (Japan) and LAPNET (Finland) arrays were selected (Fig. 5, blue triangles), as they are both dense (ca. 40 stations) and with relatively small apertures compare to their relative distance (ca. 63°). We first repeat the processing of the selection of HCDs and LCDs for this selected dataset. The comparisons between correlations stacked for the whole year for the HCDs and the LCDs are shown in Fig. 6. Specifically, we focus on S, SS and PPdf waves at long period (Fig. 6a–c), and on the

Conclusions

Nishida (2013) and Boué et al. (2013) showed that the global scale propagation of body waves can be retrieved by cross-correlation of continuous records. The conditions of the reconstruction of the deep body phases are different for the period band considered here. Two elements are to be considered: the nature of the excitation, and the part of the scattering associated to wave propagation.

The relation between seismicity and deep phase reconstructions at long period (25–100 s) is illustrated

Acknowledgments

This study was supported by the European Research Council through the advanced grant Whisper 227507. All of the data used in this study were obtained through the IRIS, NIED and RESIF data centers. We used data from numerous seismic networks, and we thank all of the dedicated seismologists and technical staff who run these networks for making their seismic data available. Finally the authors thank Kees Wapenaar and an anonymous reviewer for their careful review and constructive comments that

References (34)

  • A.M. Dziewonski et al.

    Preliminary reference earth model

    Phys. Earth Planet. Inter.

    (1981)
  • G. Ekström et al.

    The global CMT project 2004–2010: Centroid-moment tensors for 13,017 earthquakes

    Phys. Earth Planet. Inter.

    (2012)
  • M.H. Ritzwoller et al.

    Ambient noise tomography with a large seismic array

    C. R. Géosci.

    (2011)
  • E. Ruigrok et al.

    Extraction of p-wave reflections from microseisms

    C. R. Géosci.

    (2011)
  • F. Ardhuin et al.

    Ocean wave sources of seismic noise

    J. Geophys. Res., Oceans

    (2011)
  • P. Boué et al.

    Teleseismic correlations of ambient seismic noise for deep global imaging of the earth

    Geophys. J. Int.

    (2013)
  • M. Campillo et al.

    Long-range correlations in the diffuse seismic coda

    Science

    (2003)
  • B. Froment et al.

    Estimation of the effect of nonisotropically distributed energy on the apparent arrival time in correlations

    Geophysics

    (2010)
  • Y. Fukao et al.

    Seafloor topography, ocean infragravity waves, and background love and Rayleigh waves

    J. Geophys. Res. Solid Earth

    (2010)
  • J. Garnier et al.

    Passive sensor imaging using cross correlations of noisy signals in a scattering medium

    SIAM J. Imaging Sci.

    (2009)
  • J. Garnier et al.

    Correlation-based virtual source imaging in strongly scattering random media

    Inverse Probl.

    (2012)
  • P. Gerstoft et al.

    Global p, pp, and pkp wave microseisms observed from distant storms

    Geophys. Res. Lett.

    (2008)
  • G. Hillers et al.

    Global oceanic microseism sources as seen by seismic arrays and predicted by wave action models

    Geochem. Geophys. Geosyst.

    (2012)
  • S. Kedar et al.

    The ocean's seismic hum

    Science

    (2005)
  • M. Landès et al.

    Origin of deep ocean microseisms by using teleseismic body waves

    J. Geophys. Res. Solid Earth

    (2010)
  • E. Larose et al.

    Fluctuations of correlations and Green's function reconstruction: Role of scattering

    J. Appl. Phys.

    (2008)
  • F.-C. Lin et al.

    Seismic interferometry with antipodal station pairs

    Geophys. Res. Lett.

    (2013)
  • Cited by (71)

    • CCMOC: A new view of the Earth's outer core through the global coda correlation wavefield

      2023, Physics of the Earth and Planetary Interiors
      Citation Excerpt :

      An obvious significant advantage over previous approaches was that a new class of studies could be performed without earthquakes, relying merely on a constant ground motion due to the interaction among solid Earth, oceans, and atmosphere. Like in the ambient noise methods, a 2-D cross-correlation stack organized in inter-receiver distance bins can be obtained using the earthquake late coda (e.g., Boué et al., 2014; Phạm et al., 2018). This 2D representation of cross-correlation as a function of inter-receiver distance is a global correlogram (for a recent review, see Tkalčić et al., 2020).

    • Insights into the dynamics of the 2010 Eyjafjallajökull eruption using seismic interferometry and network covariance matrix analyses

      2022, Earth and Planetary Science Letters
      Citation Excerpt :

      First, the CCFs computed for different station pairs show coherent arrivals at short lag-times, i.e. with apparent velocities greater than 4 km/s (Fig. 4) consistent with body waves. Other authors have shown the interferences of teleseismic arrivals close to zero lag-time (e.g. Boué et al., 2014). Second, direct wave arrivals also appear in the acausal part of the CCFs for stations located at different distances along the same radial line from the volcano (e.g., at the beginning of April in Fig. 4).

    • The Earth's coda correlation wavefield: Rise of the new paradigm and recent advances

      2020, Earth-Science Reviews
      Citation Excerpt :

      It has become a usual practice to form stacks of cross-correlation functions with inter-receiver distance (correlograms) of both the ambient noise and the seismic-event coda wavefields (Fig. 2a). Apart from the retrieval of surface waves, many features that resemble body-wave phases in the direct wavefield were observed in those stacks for both regional (e.g., Poli et al., 2012a, 2012b; Zhan et al., 2010) and global scales (Boué et al., 2014, e.g., Boué et al., 2013; Lin et al., 2013; Lin and Tsai, 2013; Nishida, 2013; Xia et al., 2016). However, the accurate “body-wave reconstruction” requires seismic sources, which can be either primary or secondary, to be concentrated near the stationary points.

    View all citing articles on Scopus
    View full text