| dc.description.abstract | The water, which penetrated the upper part of the crust gets dragged down in a subduction zone, consecutively it gets released due to increasing heat and migrates in the above tectonic plate. This is believed to be the main cause for tremors, earthquakes and volcanism. Until now these regions were investigated mostly using tomography or full waveform inversion (fwi) on dense data gathers. A possible way to increase our knowledge of the water cycle in those subduction zones is to apply full waveform inversion even on sparse data. The goal of this thesis is to show that low velocity zones are better imaged using fwi and that sparse data gathers can still produce usable high quality results.\\
Synthetic data is generated to simulate a subduction zone with water infiltration and an integer subduction zone without any modification. The models are inspired by a study carried on in the Japan Kuril trench in 2013. By comparing simple first arrival tomography with multiple full waveform inversion (on dense, medium and sparse data) outputs, we show that a sensible increase in quality of the results is obtained by running fwi. We also highlight how a sparse data gather is enough to obtain a sound increase in precision and sharpness compared to only running tomography. Considering the computational cost and the complexity of fwi we discussed how the medium spacing data gather should be optimal for processing a seismic survey and obtaining high definition results. | |
