General algorithmic description of the AUTO module incorporated in ParkSEIS (v. 3.0) is presented here. A detailed user guide is available here. To start the automatic process, select "AUTO" item from the main menu (yellow arrow), or select "AUTO" tab from the tool panel (green arrow) as illustrated below.
analysis tools) that required the user's intervention to identify the corresponding energy pattern for the M0 trend. This step can be fairly straightforward if the trend is the only dominant energy pattern present in the dispersion image [also known as "overtone (OT)" image]. Or, as is more commonly the case, it can be a challenging task if the overtone record includes other subsurface conditions, computational artifacts due to Gibb's phenomenon and boundary effects, body wave effects, etc.
As of this version of ParkSEIS, detection of the fundamental-mode (M0) dispersion trend is now handled through a fully automated process that works on various (theoretical and empirical) aspects of dispersion image data. The process is similar to a pattern recognition technique that is grounded in the theory of wavefield transformation that generates the overtone record. The actual technical implementation is linked to a significant number of empirical parameters that are properly calibrated for the condition of the original seismic field record as well as the corresponding overtone record. This empirical calibration has been made using several hundred field data sets acquired at sites of diverse subsurface conditions.
The detection of M0 trend is made through a comprehensive evaluation of the overtone record, which evaluates the following five (5) attributes of the record:
- Pattern: Recognition of the most dominant surface wave event is assessed in its spectral and phase velocity attributes. Then, the overall strength and continuity of the pattern is estimated from the surface wave dispersion perspectives.
- Coherency: Most major energy trends are identified and their relative coherencies are estimated.
- Definition: Major energy segments are selected and then evaluated for their strengths in comparison to those of the body wave segments.
- Amplitude: Major energy segments are selected and then evaluated for their energy levels with respect to the background noise level.
- Energy: An absolute amount of surface wave energy is estimated.
The detection process accounts for all of these attributes (with varying degrees of contribution). The key parameter, however, is the "Pattern" which is calculated first and then upon which other relative contributions are determined. The detected M0 trend is used to set a specific area (called the "Bound OT") in the original overtone image to which the subsequent process of the curve extraction should focus. This is similar to setting the (lower and upper) bounds along the visually identified M0 trend,
which is the mode of operation during a manual process. A flowchart of the overall process is presented in Figure 1 below.
To briefly explain each step in the flowchart:
- A seismic data set with field-geometry encoded [*(SR)*.dat] is used as an input to the 2D wavefield transformation. A dispersion image (overtone) data set [*(Auto-OT).dat] is produced from this process.
- Each overtone record is then evaluated by a pattern recognition method for the amount of signal wavefields (i.e., M0 surface waves) with respect to all other wavefields (i.e., for SN ratio). Then, a new set of overtone data [*(Bound-OT).dat] is prepared by applying proper signal-enhancing techniques to the original overtone records [*(Auto-OT).dat].
- A special type of dispersion curve [*(BoundDC).dc] is extracted from each record in this new data set. This curve contains information about detected M0 trend in the original overtone record [*(Auto-OT).dat].
- An M0 curve is extracted from each of the original overtone records [*(Auto-OT).dat] by using the special dispersion curve from (c) as a bound curve. All extracted curves are then saved and/or passed to the next inversion process to generate a (1D or 2D) velocity (Vs) profile.