Full waveform inversion (FWI) of surface waves with genetic algorithm (GA) is able to invert complex near surface models even in the case where very limited a-priori information is available, but it requires long computing time. One way to reduce the computing time is to make assumptions on the subsurface and to simplify the forward modelling. By using a few complex near surface models, with velocity inversions, lateral velocity variations and with an irregular topographic surface, we discuss how the following issues affect the inversion results in terms of either the data misfit or the model misfit: 1) fixing the compressional wave velocities and densities to the estimated shear wave velocities according to empirical equations, instead of inverting them; 2) neglecting attenuation in the forward modelling; 3) performing 2D forward modelling and applying a 3D to 2D correction to the observed data. Although these approximations degrade model prediction, yet the main features of the shear wave models can be retrieved. Instead, the data prediction is always satisfactory, showing again that theoretical approximations in the forward modelling affect more the model misfit than the data misfit.
Surface Wave FWI on Complex Models - The Robustness of the Inversion to Assumptions and Forward Modeling Approximations
Xing, Z.
;Mazzotti, A.Supervision
2017-01-01
Abstract
Full waveform inversion (FWI) of surface waves with genetic algorithm (GA) is able to invert complex near surface models even in the case where very limited a-priori information is available, but it requires long computing time. One way to reduce the computing time is to make assumptions on the subsurface and to simplify the forward modelling. By using a few complex near surface models, with velocity inversions, lateral velocity variations and with an irregular topographic surface, we discuss how the following issues affect the inversion results in terms of either the data misfit or the model misfit: 1) fixing the compressional wave velocities and densities to the estimated shear wave velocities according to empirical equations, instead of inverting them; 2) neglecting attenuation in the forward modelling; 3) performing 2D forward modelling and applying a 3D to 2D correction to the observed data. Although these approximations degrade model prediction, yet the main features of the shear wave models can be retrieved. Instead, the data prediction is always satisfactory, showing again that theoretical approximations in the forward modelling affect more the model misfit than the data misfit.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.