Time-domain poroelastic full-waveform inversion of shallow seismic data

Among seismic imaging methods, Full-waveform inversion (FWI) is a high-resolution imaging technique to recover the geophysical parameters of the elastic subsurface from the entire content of the seismic signals. However, the subsurface material properties are less well estimated with only elastic constraints, especially for the near-surface structure, which usually contains fluid contents. Along with the deep-going research on FWI, its performance prospects on complex media with multiple parameters are urging to be investigated. On another side, Biot’s theories, which consider the stress from the solid skeleton and the fluid environment, can provide a framework to describe seismic wave propagation in the poroelastic medium.

In this thesis, a 2D time-domain (TD) Rayleigh/Love wave poroelastic FWI (PFWI) algorithm is proposed by applying the fluid-saturated poroelastic equations to carve the physical mechanism in the shallow subsurface. A FORTRAN inversion package $IforPoro$ is developed to conduct the numerical reconstruction tests. To detect the contribution of the poroelastic parameters to shallow seismic wavefields, the scattered P-SV&SH wavefields corresponding to a single model parameter are derived explicitly by Born approximation and shown numerically afterward. ... mehrThe Fréchet kernels are derived and exhibited in P-SV&SH schemes to analyze the sensitivities of the objective function to different poroelastic parameters. A series of numerical tests on gradients with respect to different model parameters are performed to evaluate inter-parameter trade-offs further. The derivations are verified by the mono-parameter PFWI reconstructions when the target model contains a porous anomaly and layered poroelastic medium. For multi-parameter PFWI tests, a cross-target model is set up with changing patterns of combined anomalies. Both Rayleigh and Love wave PFWI are numerically implemented for comparisons.

The results indicate that the fluid information (e.g., fluid density $\rho_f$, porosity $\phi$) can be recovered directly by PFWI, which behaves with strong nonlinearity. The poroelastic model parameterization should only contain rock physical properties (e.g., elastic modulus, bulk modulus, density, etc.) instead of updating effective velocities, which are strongly coupled with parameters. For different target areas, model parameters can be split into groups that get similar benefits from wave information.