Elsevier

Computers & Geosciences

Volume 135, February 2020, 104360
Computers & Geosciences

Quantification of the resolution of dispersion image in active MASW survey and automated extraction of dispersion curve

https://doi.org/10.1016/j.cageo.2019.104360Get rights and content

Highlights

  • Quantification of the resolution of dispersion image from active MASW survey.

  • Application of quantified resolution for identifying the optimal offset distance.

  • Automated extraction of dispersion curve from dispersion image.

  • Elimination of user subjectivity on the repeatability of the Vs profiles.

Abstract

Multichannel analysis of surface waves is a non-invasive method in the field of geophysics which is used for identifying shear wave velocity profile of the subsurface along its depth. The reliability and accuracy of the obtained profile strictly depends on the resolution of the dispersion image obtained from an active survey. The dispersion image is a representation of the distribution of the energy of propagating waves on a frequency-phase velocity space. Based on the relative energy accumulation in the dispersion image, the fundamental (M0) and higher modes of dispersion can be identified. As there are several parameters that can affect the resolution of dispersion images, it is important to determine the optimum magnitude of each parameter that would contribute in generating dispersion images of higher resolution. Until date, the resolution of dispersion image is described only qualitatively, wherein fundamental mode dispersion curve was extracted based on the best possible visual discretion of the user. Erroneous selections of dispersion points often led to the misidentification of the dispersion mode, leading to incorrect shear wave velocity profile. This paper reports an image-processing based methodology to quantify the resolution of dispersion images and automated extraction of M0 dispersion curve through threshold energy filtering of the dispersion image. Such methodology removes the user discretion-induced subjectivity associated with the manual extraction of dispersion curve, thereby resulting in a robust extraction technique.

Introduction

‘Multichannel analysis surface waves’ (MASW) is a non-destructive seismic survey method that used to evaluate the stiffness profile of the substrata along its depth. Owing to the portability of the equipment and ease of conducting the survey, this method proved to be an effective alternative to the conventional seismic borehole surveys. Depending on the type of source used to generate the vibrations, MASW surveys are classified as Active, Passive Roadside or Passive Remote survey. Any type of MASW survey comprises three primary procedures, i.e. data acquisition, dispersion analysis and inversion analysis. The procedure of data acquisition involves the recording of the propagating vibration wavefields using a set of geophones placed on the ground surface in a specific arrangement. Dispersion analysis reveals the distribution of propagation energy on a phase velocity-frequency space, from which the fundamental mode (M0) dispersion curve is identified and extracted. Inversion analysis is used to obtain the shear wave velocity (Vs) profile by suitably inverting the M0 curve with the aid of an optimization algorithm. The success of the entire process in obtaining a meaningful and reliable Vs profile is largely dependent on the resolution of dispersion image. This paper reports about quantifying the resolution of dispersion image obtained from an active MASW survey, and subsequent automated extraction of fundamental mode dispersion curve (M0).

Dispersion is the phenomenon associated with the waves propagating through homogeneous or stratified soil layers, where each of the frequencies travel with a specific or multiple phase velocities. The collected wavefield records are preprocessed using temporal muting and frequency filtering. The processed record is subjected to dispersion analysis to establish a relationship between the frequency (f) and phase velocity (C) of the constituent waves (Taipodia et al., 2018a). It is to be noted that transferring the time domain data in the frequency domain has an inherent assumption of linear response of the system. Since the strains associated with the generated vibrations particularly belong to the extremely low strain problems, the response of the soil remains linear. Thus, the application of FFT is completely acceptable and conforms to the linear characteristics of the soil system. Unlike the earlier methods of direct estimation of group-velocity or apparent phase-velocity dispersion curves (Jin and Colby, 1991; Park et al., 1998), the multichannel approach does not estimate the individual phase velocities, rather constructs an image space where the dispersion trends are represented by the energy accumulation pattern in the f-C domain. Fast Fourier Transform (FFT) is generally adopted to obtain the dispersion image. All types of propagating seismic waves, possessing considerable energy, are eventually imaged. From the dispersion image, the pattern in energy accumulation is visually identified, and the trend comprising the peak energy is manually extracted. As shown in Fig. 1, depending on the presence of single or multiple phase velocities for a particular frequency, or vice-versa, the extracted dispersion modes can be unimodal or multimodal, respectively (Park et al., 1998, 2007).

In any dispersion imaging scheme applied to experimentally recorded wavefield, the normalized energy of the propagating waves is mapped on the dispersion image space. Theoretical dispersion curve, obtained from any of the standard methods (Schwab and Knopoff, 1972; Ke et al., 2011; Sadeghisorkani et al., 2008), is optimized towards the experimental dispersion curve, through a sequential and iterative model parameter updating scheme. The optimization process of matching the theoretical dispersion curve, obtained from a forward analysis with an assumed soil stiffness profile, against the experimental dispersion curve is usually termed as inversion analysis. A common practice is to reduce the root-mean-square (RMS) difference between the theoretical and experimental dispersion curve to a minimal. As the tolerance is satisfied, the optimized theoretical dispersion curve is accepted to be the final dispersion model to be used in inversion analysis. In recent decades, there have been successful applications of bio-inspired optimization algorithms in the inversion of dispersion curves (Corchete, 2012; Kulesh et al., 2008; Song et al., 2008, 2009, 2012, 2015). However, the efficacy of the inversion analysis depends on the precision of extracting the M0 dispersion curve which is dependent upon the resolution of the corresponding dispersion image. The effects of various data pre-processing and data acquisition parameters on the resolution of the dispersion image are already reported in various literatures (Park et al., 2002; Zhang et al., 2004; Dikmen et al., 2010; Taipodia et al., 2017, 2018b). However, in all the reports, based on visual inspection, the resolution of the images was merely described qualitatively, while the quantification of the resolution was completely omitted.

Commercially available software, such as Surfseis, Geopsy or EasyMASW, permits only a user controlled manual extraction of dispersion curve from the dispersion image space. Such approach often leads to erroneous extraction due to the indistinct identification of the precise location of the locally highest peak energy in the dispersion image. Any small error in locating the peak energy would automatically choose an erroneous dispersion point with lower energy. This leads to the selection of a dispersion curve that is misread to be the fundamental mode. Such erroneous selection, when subsequently used in inversion, results in an inaccurate estimation of the shear wave velocity profile. Hence, it is extremely necessary to develop a methodology free from user subjectivity that can be used to unambiguously locate the locally highest energy peaks.

This paper describes a strategy to address the above two issues, namely (a) quantifying the resolution of dispersion image, and (b) automated extraction of precise M0 curve from the dispersion trends. In this regard, several active MASW surveys is carried out at the campus of Indian Institute of Technology Guwahati, Assam, India. The generated wavefields were collected, the data were processed and the corresponding dispersion image was generated using SurfSeis, the post-processing software. The tests were conducted using varying offset distances, and the influence of the latter is exhibited on the resolution of corresponding dispersion images. In order to quantify the resolution of dispersion image and automated extraction of dispersion curve, customized MatLab codes are integrated with the Surfseis post-processing simulations, wherever required. The following sections provide a description of the various steps associated with the developed methodology.

Section snippets

Active MASW experimentation program

Field investigation utilizing active MASW survey was carried out at the campus of Indian Institute of Technology Guwahati, India, specifically at the location of the Block-D residential apartments (26°11′05.4″N, 91°41′31.8″E) as shown in Fig. 2(a). According to the borehole stratigraphy reports, the site consists of a shallow stiff soil layer of 7 m, underlain by a hard stratum. Fig. 2(b) shows the representative borelog as obtained for the site. It is worth mentioning that the constructions at

Influence of offset distance on the resolution of the dispersion image

Planar wave propagation is effective only after the wave travels a certain distance from the source (Stokoe et al., 1994; Park et al., 1999). There are two kinds of effects due to offset distance, namely the near-field effect and the far-field effect (Park et al., 1999; Park, 2011). The unpredictable non-planar propagation of surface waves near the source is referred as near-field effect. Non-planar propagation of waves may lead to the interference of multiple reflections and mode conversions

Quantification of the resolution of dispersion image

Various researchers have described the resolution of dispersion image on a qualitative basis, which led to the extraction of M0 dispersion curve subjected to user discretion (Park et al., 1998; Moro et al., 2003; Dikmen et al., 2010; Zhang et al., 2004; Taipodia et al., 2018b). This paper provides a basis of quantifying the resolution of dispersion image obtained from active MASW surveys. The resolution of an image is defined by the numbers of pixels contained in a unit area, which quantifies

MATLAB based image processing to quantify resolution

Image Processing Toolbox provides a comprehensive set of reference-standard algorithms and workflow apps for image processing, analysis, visualization, and algorithm development. Image segmentation, image enhancement, noise reduction, geometric transformations, image registration, and 3D image processing can be efficiently performed using the toolbox command menu. A MatLab coding (DISPBANDJASD.m) is developed to provide a quantification of the resolution of dispersion image in terms of the

Automated extraction of dispersion curve from dispersion image

It is well understood from various literatures (Park et al., 1998, 2002; Zhang et al., 2004; Xu et al., 2006; Luo et al., 2008; Taipodia et al., 2018c, Taipodia et al., 2018b, Taipodia et al., 2018a) that depending on several geometrical and physical parameters, the dispersion image can be fuzzy and obscure. Dispersion curve, as stated earlier, represents the distribution of localized highest energy over a wide range of phase velocity and frequency. The energy accumulations are represented as

Summary and conclusions

The present study highlights about quantifying the resolution of the dispersion image obtained from the dispersion analysis using SurfSeis. A MatLab code, encompassing image processing techniques, has been developed to characterize the resolution of dispersion image in terms of the thickness of dispersion band, its frequency coverage and continuity over the frequency band. It is illustrated that the quantification of the dispersion image resolution can be instrumental in justifiably identifying

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

The authors sincerely acknowledges the reviewers whose critical comments and suggestions have helped in significant improvement in the quality and clarity of the article.

References (26)

  • T. Imai et al.

    Correlation of N-value with S-wave velocity and shear modulus

  • G. Ke et al.

    Modified Thomson–Haskell matrix methods for surface-wave dispersion-curve calculation and their Accelerated root-searching schemes

    Bull. Seismol. Soc. Am.

    (2011)
  • Y. Luo et al.

    Rayleigh-wave dispersive energy imaging using a high-resolution linear radon transform

    Pure Appl. Geophys.

    (2008)
  • Cited by (0)

    1

    Dr. Taipodia is associated with the experimental investigation, data processing, image analysis and preparation of the manuscript.

    2

    Dr. Dey has been involved with the overall supervision and arrangement of the resources for the project as well as data analysis.

    3

    Dr. Gaj has been associated with the development of Matlab codes required for the image processing.

    4

    Mr. Baglari has been associated with the field experimental investigations for the project.

    View full text