Despite its apparent advantage over the two-channel SASW (spectral analysis of surface wave) method, the testing configuration of the MASW (multi-station analysis of surface wave) method remains a crucial factor that may affect the test results. Tradeoffs are involved when selecting the testing parameters, In addition, several algorithms with different preference in the literature exit for the dispersion analysis. The objectives of this study are to establish a standard procedure for field testing and dispersion analysis of MASW.
In the part related to field testing, the influences of temporal and spatial parameters are investigated, which includes aliasing and leakage in both time and space domain, far and near field effects, higher mode domination, and horizontal resolution. The investigation leads to several rules for choosing testing parameters. An innovative testing procedure, called the pseudo-section method, and the associated signal processing is proposed to resolve the dilemma of choosing testing parameters and standardize the testing procedure. In addition, a complementary seismic source with greater energy and lower frequency contents and a non-invasive receiver stream are developed and evaluated in this study. Major conclusions drawn from this part of study include:
1. The field testing basically involves generating a surface wavefield and recording discretized wavefield. The discretization and truncation in the time and space domain result in aliasing and leakage in the spectral analysis. While modern seismograph possesses sufficient sampling rate and recording time, the sampling of the space data by geophones is relatively limited. The geophone interval and spread length respectively determines the shortest and longest wavelengths that can be analyzed.
2. The effects of multiple modes on multi-station measurements are investigated and the criterion of mode separability is discovered. The offset range (conventional geophone spread length) required to separate two modes is inversely proportional to the difference in wave number. A long geophone spread is required to have great depth coverage and separate potential higher modes, but is unfavorable for lateral resolution.
3. The near offset (nearest source-to-receiver distance) determines the offset range affected by near and far field effects. Mitigations of the near field and far field effects by choosing appropriate near offset are in conflict.
4. The dilemma in deciding the geophone spread and near offset can be resolved by the proposed pseudo-section method. It consists of a walk-away survey and a phase-seaming procedure when synthesizing seismograms with different nearest source-to-receiver offset, allowing wide-wavelength dispersion analysis within a small spatial range.
5. The standard field configuration is proposed. The geophone interval and near offset is determined by the shortest wavelength of interest and the actual geophone spread depends on interested spatial resolution. The number of walk-away shots should be enough for the longest wavelength of interest and separating possible higher modes.
To improve testing efficiency, the source and receiver may be exchanged and a non-invasive receiver stream can be used. For long offset, a heavy weight-drop source can be used to enhance low frequency measurements.
In the part of dispersion analysis, the procedures for signal quality assessment and a unified approach for wavefield transformation are proposed. Furthermore, the differences between conventional wavefield transformations and data sampling of dispersion curve are
discussed. The major conclusions are illustrated as follows:
1. The discrete wavefield in time-space (t-x) domain is transformed to frequency- space (f-x) domain first. The information revealed from the complex data in the f-x domain can be helpful on signal quality controls and preliminary understanding of results, which includes:
The energy spectrum provides indication of effective frequency range and the variation of phase angle with distance allows preliminary dispersion analysis (a new method called multi-channel spectral analysis of surface wave, MSASW).
In addition to the preliminary dispersion curve, MSASW can also evaluate the data quality because it is based on the linear regression of phase angles measured at multiple stations. The proper range of offset for constructing the dispersion curve can be selected, and existence of multiple modes may be identified.
The real part of f-x complex data is shown to be useful for optimum offset selection to reduce the effect of near and far field effects.
2. The f-x domain is further transformed to f-k (wavenumber), f-p (slowness), f-v (velocity), or f-λ (wavelength). The f-k transformation, f-p transformation, and phase shift methods were done with different algorithms and conclusion of their performance comparison varies in previous studies. In this study, a new unified algorithm is proposed that transforms the time-space wavefield into f-k, f-p, f-v and f-λ domains simultaneously.
3. The dispersion curves extracted from different domains may be different because the discretization in different domain implies different resolution in the dispersion analysis. Although the recorded time-space wavefield is discrete, the 2D wavefield transformation remain continuous in theory (i.e. f, k, p, v, and λ are continuous
quantities). The dispersion curves obtained by different transformation are shown to be identical by a newly-proposed optimization method based on the discrete-space Fourier Transform, which allows the transformed domain remain continuous for best resolution of dispersion analysis.
4. This study further investigates the data sampling of the dispersion curve. The wave propagation theory shows the wavelength (λ) has a close relation to the influence depth, nevertheless conventional wavefield transformations never directly deal with it. The conventional dispersion analysis samples equally in f domain and causes over sampling in short wavelengths (shallow depths). A wavelength-controlled sampling approach is proposed for the dispersion curve to avoid bias in depth sampling.