Work

Surface Wave Tomography at Exploration Scale

Year: 2013
Type: dissertation
Abstract: In this work we present the application of a direct tomographic inversion method, similar to that historically used in global seismology, to exploration scale. We will start with an excursus on the physical background necessary to understand how surface waves propagate and the charachteristics which make these waves suitable for inversion. In the following section the optimal acquisition methods will be described. These are not very different from those for the datasets used for industry-standard reflection seismology imaging; however a few remarks have to be made. For instance, the length of the line, the spacing between receivers and the dominant wavelength are essential parameters to take into account when acquiring a dataset for surface wave analysis. The following section, regarding the processing, will present the most used methods for inversion of surface waves at exploration scale, those at regional (seismological) scale and the method used to perform the inversion. These can be classified into tomographic and non tomographic approaches. In both cases, the aim is to extract phase velocity dispersion curves from the data and invert them to get 1D models. The key difference lies in how the dispersion curves are extracted and the "meaning" we give to them. When performing a non tomographic inversion it is possible to extract the curves in many ways, e.g. by computing a Fourier transform in space and time (f-k domain) and pick the curve as energy maxima, as surface waves are the most energetic event in a seismogram. Most commonly, the f-k spectrum is an average one over a certain number of stations. Each of these curves will be inverted to get a 1D model and each of these models can be connected to the others via constraints (regularisation). On the other hand, when one wants to perform tomographic inversion, the dispersion curves are extracted by computing the average slowness of surface waves travelling along a path from one receiver to another, illuminated by the same source. Therefore, since more paths can cross the same model point, more dispersion curves have to fit the model parameters at each model point. Seismologists, at this point, generally compute phase velocity maps from which they extract further dispersion curves to be inverted. What's new in the method used for this work is the fact that we invert dispersion curves, without computing phase velocity maps. This is why it's called "direct inversion". This has several positive aspects, e.g. the computational efficiency. Another advantage, common to all tomographic approaches is the fact that it uses only two stations and therefore does not need a dataset with a large number of channels. On the other hand, it has some limitations both on the acquisition parameters (especially offset and distance between the two receivers) as well as the fact that it only inverts for the fundamental mode dispersion curve. Two datasets have been analysed: one shot in New Zealand by ETH Zurich, one shot in the area of Torino by the Italian National Centre for Research (CNR). The first constitutes of five 2D seismic lines, one of which at ultra-high resolution (1m receiver spacing and 2 m source spacing). This dataset has already been inverted using the estabilished multichannel analysis of surface waves method. Therefore, it was a good dataset to start with as results could be compared with those from this previous inversion. The CNR dataset is a small 3D dataset, above an artificially made tank filled with loose sand. The fact that the map of the subsurface was already known made it a good dataset to test the validity of the method also in 3D. In the appendix, we show tests performed on a large 3D seismic survey, acquired in Oman for oil exploration purposes. This dataset is known in the geophysical community (e.g. Gouédard et al., 2012) for having a low signal to noise ratio. Furthermore, the subsurface to be mapped shows strong lateral variations. For this reason, preliminary tests were performed in order to find a method to increase the signal to noise ratio and therefore assess the feasibility of a direct tomographic inversion. A good solution could be to stack the traces from two different sources aligned with the same couple of receivers.
Author: F. Da Col
Language: English
Subfield: Geophysics
Open Access status: green
OpenAlex ID: W46717596
Cites: 23
Cited by: 2
Related to: 9