Spectral element elastic wave modelling of Distributed Acoustic Sensing data for the analysis of limited broadside sensitivity, in a surface seismics configuration.
MetadataShow full item record
Distributed acoustic sensing (DAS) uses fibre optic cables to measure or monitor strain in the subsurface. With DAS being a relatively new technique it gained recently more interest for geophysical studies. The main advantages of DAS are; its high spatial resolution which is between 1-10 m, its successful use in high- temperature conditions and the low costs compared to the traditional geophones used. Eventually DAS can be used for monitoring geothermal projects and CO2- sequestration projects (Wang et al., 2018). With distributed acoustic sensing the strain or strain rate is measured, based on the Rayleigh scattering principle. Pulses of laser signals are sent through the fibre and the interrogator measures the back-scattered light of these pulses. The time of the laser signal and it's back-scattered component are measured and converted to strain, using the speed of light and the refractive index of the fibre (Jousset et al., 2018). One of the challenges that DAS faces is measuring the P-waves that enter the fibre with a large angle, this is called limited broadside sensitivity. When the fibre is orientated parallel to the surface, the fibre can be insensitive to P-waves travelling vertically. Kuvshinov (2016) suggest a correction factor for this lim- ited broadside sensitivity, which will be tested in this thesis. With the advantages and challenges of DAS, the focus of this thesis will be on limited broad side sensitivity and the main question of this master thesis is as follows: Can reflection information from Distributed Acoustic sensing be improved for broadside P-waves, in a surface seismics configuration? The approach for this new technique, is to first model DAS data with the spec- tral element solver SPECFEM. SPECFEM is an open-source spectral elemental modelling program, which can be used to model wave propagation (particle velocity). The aim is to translate the synthetic particle velocity data to DAS (strain) data, in order to be able to understand DAS data completely and be able to compare it to real DAS data. From the synthetic particle velocity it will be researched if broadside waves can be captured. Which is one step towards achieving synthetic DAS data. In order to research the limited broadside sensitivity issue, the steps that need to be taken are: filtering out the surface waves, selecting the peak amplitude of the P-wave reflection and applying the correction factor for broadside sensitivity cos^2(θ)/sin(θ) (θ being the angle between the incoming and the horizontal fi- bre.), obtained from two point ray tracing. By applying this factor the synthetic DAS data is simulated. Then it can be tested if the reversed correction factor sin(θ)/cos2(θ) can correct the simulated DAS data for the limited broadside sensitivity, so that it can be interpreted. Based on the results it can be concluded that SPECFEM can be used success- fully for modelling DAS configurations. The CREWES toolbox from MATLAB is a sufficient tool box for ray tracing and achieving travel time and horizontal slowness. By applying a slope filter, to extract the surface waves, it became clear that a receiver spacing of 0.5 m is preferred for the model considered here. Furthermore with the use of the correction factor for broadside sensitivity, DAS data can successfully be simulated and the reversed correction factor can be used to correct for limited broadside sensitivity in DAS data. The limited broadside sensitivity is only one step in modelling DAS data and achieving synthetic DAS data. For further research it is desirable to improve the isolation of the P-wave reflections from the subsurface interface of inter- est, by designing a more advanced amplitude picker. To get a more realistic model, noise should be added to the reflection data followed by stacking the reflection data. SPECFEM3D instead of SPECFEM2D, is recommended to be used for further modelling , because of a more realistic 3D geometrical spread- ing of the source. For modelling DAS data the next step would be to convert the particle velocity to strain and research other parameters to eventually be able to compare it to real DAS data. But it is shown that DAS looks like a promising technique, where the outcomes of the considered model shows the correction factor for broadside sensitivity can improve the data for relatively small source-receiver offsets (x=-250,+250 m).