High resolution imaging of lithospheric structures by full waveform inversion of short period teleseismic P waves

Seismic tomography allows us to image the Earth’s interior based on surface observations of seismic waves. The full waveform inversion (FWI) method has the potential to improve tomographic images for the fine scale structures of the lithosphere. For this reason it receives a lot of attention of seismologists. FWI requires an efficient and precise numerical techniques to solve the elastic wave equation in 3D heterogeneous media. Its resolution potential is limited by the shortest wavelength in the seismic wavefield and the wavefield sampling density. Because of the high computational cost of modeling the propagation of seismic waves in heterogeneous media, FWI remains challenging. However, owing to the progress in high performance computational resources and numerical simulation techniques, as well as the deployment of permanent and temporary broadband arrays in the last two decades, this situation has changed dramatically. In this thesis, we focus on the high resolution imaging of lithospheric structure beneath the Pyrenean range by FWI, to quantify the highly controversial amount of convergence that occurred during the formation of this mountain range. In order to obtain finely resolved tomographic images, we exploit short period teleseismic P waves recorded by dense transects. We use a hybrid method that couples a global wave propagation method in a 1D Earth model to a 3D spectral-element method in a regional domain. A boundary coupling approach is used to match the global and regional wavefields on the boundaries of the regional domain. This hybrid method restricts the costly 3D computations inside the regional domain, which dramatically decreases the computational cost. The hybrid method can model teleseismic wavefields down to 1s period, accounting for all the complexities that may affect the propagation of seismic waves in the 3D regional domain. By using this hybrid method, the sensitivity kernels of the least square waveform misfit function with respect to elastic and density perturbations in the regional domain are computed with the adjoint state method. These waveform sensitivity kernels are used in an iterative L-BFGS algorithm to invert broad-band waveform data recorded by two dense transects deployed during the temporary PYROPE experiment across the Pyrenees mountains. We obtain the first high resolution lithospheric sections of compressional and shear velocities across the Pyrenean orogenic belt. The tomographic models provide clear evidence for the underthrust of the thinned Iberian crust beneath the European plate and for the important role of rift-inherited mantle structures during the formation of the Pyrenees.