Previous studies indicate that scattering may pose a trade-off for the performance of seismic interferometry (SI) applications for retrieving body-wave reflections of a target reflector. While it has been demonstrated that a higher scattering strength of the overburden improves the Green's function estimated by cross-correlation SI, other theoretical and empirical studies showed that multiple scattering also gives rise to more artefacts. The implications of this trade-off are analysed in this numerical study for a lithospheric scenario with varying crustal scattering strength and passive illumination conditions. In this scenario, we apply SI by cross-correlation to elastodynamic responses to double-couple sources to reconstruct virtual Moho primary reflections. We include multidimensional deconvolution (MDD) methods in the analysis to investigate whether scattering-induced artefacts affect MDD methods in a similar way as was shown for the cross-correlation method. Our results show that there indeed exists a trade-off between the quality of the virtual primary reflection of the target that can be obtained by SI and the scattering strength of the overburden. Furthermore, we find that the full-field MDD method proves to be most resilient to the negative effects of multiple scattering for all illumination conditions and scattering strengths analysed.

@en

Virtual Green's functions obtained by seismic interferometry (SI) can provide valuable reflectivity data that can complement tomographic inversion schemes. However, virtual reflections are affected by illumination irregularities that are typical of earthquake-induced wavefields recorded by the receiver array. As a consequence, irregular source distributions, scattering (in case of suboptimal illumination), and complex source mechanisms can significantly disturb the retrieval of Green's function approximations by conventional SI methods. We introduce SI by full-field multidimensional deconvolution (MDD) for elastodynamic wavefields as an alternative method to obtain body wave Green's functions under those typical circumstances. The advantage of this method compared to other MDD methods is that the kernel of its governing equation is exact. This alternative formulation of the kernel pertains to several advantages: the solution is less sensitive to artifacts and utilizes the free-surface multiples in the data to estimate primary reflections. Moreover, the point spread function of the full-field MDD method corrects more affectively for irregular illumination because it also addresses irregularities caused by scattering inside the medium. In order to compare full-field MDD to existing SI methods, we model synthetic earthquake recordings in a subduction zone setting using an elastodynamic finite-difference scheme with double couples of different orientations and peak frequencies. Our results show that SI by cross correlation suffers most under these circumstances. Higher-quality reflections are obtained by the MDD methods, of which full-field MDD involves the most stable inversion, and its results are least contaminated by artifacts.@en

Passive seismic interferometry enables the estimation of the reflection response of the subsurface using passive receiver recordings at the surface from sources located deep in the Earth. Interferometric imaging makes use of this retrieved reflection response in order to study the subsurface. Successful interferometric imaging relies on the availability of passive recordings from sufficient sources in the subsurface. Ideally, these sources should be homogeneously distributed, which is unlikely to happen in practical applications. Incomplete source distributions result in the retrieval of inaccurate reflection responses, containing artefacts which can disturb the interferometric imaging process. We propose an alternative imaging method for passive data based on illumination diagnosis and directionally constrained migration. In this method, passive responses from single transient sources are cross-correlated individually, and the dominant radiation direction from each virtual source is estimated. The correlated responses are imaged individually, thereby limiting the source wavefield to the dominant radiation direction of the virtual source. This constraint enables the construction of accurate images from individual sources with a significantly reduced amount of migrated interferometric artefacts. We also show that the summation of all individual imaging results improves the subsurface image by constructive interference, while migrated crosstalk and artefacts experience cancellation. This process, called Image Interferometry, shows that in case of limited subsurface illumination the interferometric integration can be applied in the image domain rather than in the virtual reflection-response domain, thus eliminating the need for the retrieval of the reflection response as an intermediate step.@en

Seismic interferometry (SI) for body waves offers the opportunity to utilize highfrequency scattering coda from local earthquakes to obtain a detailed reflectivity image of the lithosphere. In this thesis it is demonstrated that classical SI methods are seriously affected by circumstances that are typical of field data and that multiple scattering poses a complex trade-off for SI performance. Therefore, we propose an alternative method by multidimensional deconvolution (MDD) that proves to be more resilient under realistic circumstances and properly utilizes the scattering coda: full-field MDD. The main advantage of this method over classical MDD methods is that the kernel of its governing equation is exact, which allows for an optimal use of the multiple scattering coda to obtain virtual primary reflections of the lithosphere.@en

Our objective is to complement lithospheric seismic tomography with an interferometric method to retrieve high-resolution reflectivity images from local earthquake recordings. The disadvantage of using local earthquakes for the retrieval of reflected body-waves is their usual sparse distribution. We propose an alternative formulation of passive seismic interferometry by multidimensional deconvolution (MDD) which uses the multiples in the full recordings to compensate for missing illumination angles. This method only requires particle-velocity recordings at the surface from passive transient sources and retrieves body-wave reflection responses without free-surface multiples. We conduct an acoustic modelling experiment to compare this formulation to a previous MDD method and Green’s function retrieval by crosscorrelation for different source distributions. We find that in the case of noise-contaminated recordings obtained under severely limited and irregular illumination conditions, the alternative MDD method introduced here still retrieves the complete reflection response without free-surface multiples where the other interferometric methods break down. @en

Several seismic investigations - using receiver-function methods as well as tomographic approaches - have been carried out in the Malargüe region (Argentina) for various purposes over a few decades. We use a body-wave seismic interferometry (SI) approach to retrieve reflections later used for the consecutive imaging of the subsurface. We investigate the applicability of the body-wave SI using P-wave coda from local earthquakes with the aim to retrieve reflection responses from a part of the Andean crust below the seismic array we use. We called our technique local-earthquake P-wave coda (LEPC) SI. In this presentation, we show three different LEPC SI results based on three different SI theories: crosscorrelation, crosscoherence, and multidimensional deconvolution.We find that, from a structural-interpretation point of view, multidimensional deconvolution based on the truncated singularvalue decomposition scheme provides us with a better structural imaging than the other SI approaches.We interpret deep thrust faults in the imaging results from LEPC SI, whose presence in this region has previously been indicated from interpretation of active seismic-survey data and exploration-well data. We also interpret dimmed-amplitude parts in the reflection image as possible melting zones that have been previously indicated by magnetotelluric methods. The LEPC SI method we propose could be used as a low-cost alternative to active-source seismic surveys for imaging and monitoring purposes of deeper geothermal reservoirs, e.g. in enhanced geothermal systems where the target structures are down to 10 km depth. @en

Seismic interferometry (SI) presents a set of inexpensive and noninvasive methods that can be applied to any array at the surface to retrieve virtual body-wave reflection responses from earthquake recordings. Conventional SI by cross-correlation requires recordings of wavefields in lossless media generated by a smooth continuous distribution of passive sources with isotropic source radiation patterns and similar power spectra. These conditions are unlikely to be met in the lithosphere: earthquakes are distributed sparsely and generated by complex mechanisms. The resulting anisotropy in the illumination of the receiver array causes the retrieved virtual-source radiation patterns to be irregular, leading to artifacts which can obscure the desired body-wave reflections. SI by multidimensional deconvolution (MDD) can inherently correct for anisotropic illumination of the array and does not rely on the medium being lossless. We propose an alternative formulation of MDD for two-way wavefields: full-field MDD. Different from previous MDD methods for passive two-way wavefield recordings, full-field MDD uses multiples in the passive data to construct the reflection response without free-surface interaction. Therefore, this MDD method profits from additional wavenumbers provided by scattering to compensate for sparse earthquake distributions. Besides, this method does not require wavefield decomposition, which is sensitive to velocity variations at the receiver level. We compare the reflection retrieval by full-field MDD and cross-correlation for a limited passive source distribution in a lithospheric model with a discontinuous Moho at a depth of 50 km. We simulate earthquakes generated by dipole sources along a listric fault-system with power spectra varying within bandwidth 0.2-2.6 Hz. The reflection response retrieved by full-field MDD shows a continuous high-resolution Moho reflection, while cross-correlation yields a very low resolution response obscured by artifacts. @en

We discuss a method to interferometrically retrieve the body wave reflection response from local high-frequency scattering coda wave fields with the purpose to obtain an input dataset suitable for the application of advanced exploration-type imaging methods@en