Data analysis of a dedicated 3D survey requires special steps of shot/receiver (SR) setup and data rearrangement into 3D common-mid-point (CMP) gathers. Location information of SR patterns be should be able to be entered using either Cartesian (*.txt) or GPS (*.pgx) coordinate files. Subsequent data sorting into CMP gathers should take place based on the applied SR patterns by simultaneously considering a bin size and the 3D ground volume. To simplify this complicated process, it would be desirable for the numerical analysis part of dispersion and inversion to proceed in a fully automated manner to produce a 3D grid data set (*.G3D) at the end. In ParkSEIS-3D, the numerical analysis parts proceed in a fully automated manner by clicking one button as illustrated below.
There are two steps in numerical analysis of MASW survey data: dispersion and inversion.The purpose of dispersion analysis is to extract a fundamental-mode (M0) [or an apparent mode (AM0), as explained below] dispersion curve from dispersion image of one field record. The following inversion analysis is fairly straightforward and automatically generates a 1D velocity (Vs) profile by using the measured M0 curve. Correct extraction of M0 curve is therefore critically important directly influencing the accuracy of the velocity profile. The proper extraction, however, requires operator's ability to accurately interpret energy patterns observed in the dispersion image, which often requires theoretical knowledge and significant observation experience with diverse field data sets. This is why all commercial software packages require operator's manual intervention for this step
because complex multiple energy trends almost always occur originating not only from (source and non-source generated) multi-modal surface waves but also from body waves (e.g., refraction and reflection).
ParkSEIS-3D has incorporated an automatic algorithm (AUTO) similar to artificial intelligence (AI) that detects the correct M0 pattern in a mixture of complicated energy trends. The AUTO algorithm is more an art than a science because there are no well-established systematic approaches that can determine the M0 trend among highly unpredictable energy patterns. The algorithm evolved through extensive self-learning process by using diverse field data sets over the past two decades.
ParkSEIS-3D uses two approaches for inversion of the measured dispersion curve; (1) fundamental-mode (M0) and (2) apparent-mode (AM0) methods. Although there has been a great deal of research and development in the higher modes inversion, software that takes full advantage of multi modes while efficiently handling all the associated complications has not yet been developed, at least to the level of commercial software package. This is because of the modal characteristics of surface waves that are ultimately determined by the velocity (Vs) profile of subsurface materials, the unknown that we attempt to know through the MASW analysis. On the other hand, energy of the fundamental-mode (M0) surface waves almost always dominates the measured seismic wavefields. In this sense, the traditional approach of the fundamental-mode (M0) inversion provides
an excellent outcome under most common overburden/bedrock settings. On the other hand, it is well known that a velocity inversion, a layer of higher velocity (Vs) than the layer below, often results in an abnormal dispersion trend where phase velocities increase (instead of decrease) with frequencies. This trend arises from surface waves jumping from one mode to next mode where the two dispersion curves get sufficiently close to each other. The resulting trend usually appears to be one coherent mode a lot like the fundamental mode, and therefore is called an apparent mode (AM0). Existence of AM0 trend is automatically detected in ParkSEIS-3D and a different inversion method is used to account for the mode jump phenomenon. In consequence, the inverse-velocity layers are more accurately handled than by the traditional approach.
The algorithmic flowchart of ParkSEIS-3D displayed above shows the AUTO algorithms adopted in the general procedures from both conventional multiple 2D/1D surveys and a dedicated 3D survey.