Improving the quality of empirical Green's functions, obtained by cross-correlation of high-frequency ambient seismic noise

Abstract

Abstract. Studying the uppermost structure of the subsurface is a necessary part of solving many practical problems (exploration of minerals, groundwater studies, geoengineering, etc.). The practical application of active seismic methods for these purposes is not always possible for different reasons, such as logistical difficulties, high cost of work, and a high level of seismic and acoustic noise. That is why developing and improving passive seismic methods is one of the important problems in applied geophysics. In our study, we describe a way of improving the quality of empirical Green's functions (EGFs), evaluated from high-frequency ambient seismic noise, by using the advanced technique of cross-correlation function stacking in the time domain (in this paper we use term “high-frequency” for frequencies higher than 1 Hz). The technique is based on the global optimization algorithm, in which the optimized objective function is a signal-to-noise ratio of an EGF, retrieved at each iteration. In comparison to existing techniques, based, for example, on weight stacking of cross-correlation functions, our technique makes it possible to significantly increase the signal-to-noise ratio and, therefore, the quality of the EGFs. The technique has been tested with the field data acquired in an area with a high level of industrial noise (Pyhäsalmi Mine, Finland) and in an area with a low level of anthropogenic noise (Kuusamo Greenstone Belt, Finland). The results show that the proposed technique can be used for the extraction of EGFs from high-frequency seismic noise in practical problems of mapping of the shallow subsurface, both in areas with high and low levels of high-frequency seismic noise.

Deep structure of the Verkhnetovskaya kimberlite pipe in the Arkhangelsk diamondiferous province according to passive seismic and radiological methods

“…In our study, we considered the joint use of several methods: microseismic sounding (MSM) (Gorbatikov et al., 2008, 2013), passive seismic interferometry with an advanced stacking method (Afonin et al., 2019), the H/V spectral ratio method (H/V method) (Nakamura, 1989), emanation mapping (Magomedova & Udoratin, 2016) and gamma spectrometry (Babayants et al., 2006). The advantage of this set of methods is that it allows one to study the main prospecting features of kimberlite pipes, which are subvertical heterogeneities, the most contrasting horizontal boundaries, and geochemical anomalies.…”

“…Shapiro et al., 2005; Yang et al., 2007; Lin et al., 2007) and body wave imaging (Poli et al., 2012) for the study of fault structure (Afonin et al., 2017) and shallow subsurface structure (Lin et al., 2013; Cheng et al., 2015; Le Feuvre et al., 2015). In our study, we used an improved method of passive seismic interferometry, which allows for decreasing time of measurement and improving quality of the obtained dispersion curve (Afonin et al., 2019). The field measurements require simultaneous records of ambient seismic noise at least in three points.…”

Near surface structure of Sodankylä area in Finland, obtained by advanced method of passive seismic interferometry

“…However, special analysis of the continuous data would be necessary, in order to extract the diffuse wavefield from the data. For this purpose, the SNRS algorithm described earlier in Afonin et al (2019) can be used.…”

“…Therefore, empirical Greens function of the studied medium can be retrieved by crosscorrelation of this wavefield recorded at different locations (Wapenaar, 2004;Wapenaar and Thorbecke, 2013;van Manen et al, 2005). For evaluation of empirical Greens functions, we apply advanced method of passive seismic interferometry (Afonin et al, 2019) that we called signal-to-noise ration stacking (SNRS). We show results of application of passive seismic interferometry for mapping the uppermost crust in the area of XSoDEx project.…”

Automatic Image-Based Event Detection for Large-N Seismic Arrays Using a Convolutional Neural Network

“…The other demonstrated solutions for data selection are mostly semi-automatic and/or performed on preprocessed and cross-correlated data, which in turn requires extra operator workload and computational cost. Examples of these methods are based on instantaneous phase coherence [36], statistical time-series classification [37], illumination diagnosis [38], asymmetry of CCFs [39], local similarity function [40], or on signal-to-noise ratio (SNR) criterion [41][42][43][44].…”