Efficient Wavefield Modeling And Waveform Inversion For Media With Irregular Surface

Full waveform inversion(FWI)has shown great potential in imaging subsurface structures and is expected to play a more and more important role in building accurate3-D models across various length scales.To date,FWI has achieved numerous successes on marine seismic data applications,while its successful applications on land seismic data are rare,largely because land data are often acquired on an irregular surface with a relatively low signal-to-noise ratio(S/N).In this thesis,several methods and techniques are developed for effective implementation of FWI on land seismic data.The achievements in this work are summarized as follows:(1)An improved “immersed boundary” method is proposed to simulate the acoustic wavefield with irregular topography.Land seismic data are typically recorded using geophones plugged on the irregular surface.Accurately simulating the seismic wavefield with the irregular free surface is the foundation for seismic wavefield imaging and waveform inversion.With the improved “immersed boundary” method,the wavefield at the mirror points,which are at fractional grid points,is obtained through iterative symmetric interpolation.This method can also obtain the particle vibration velocity components at the free surface as seismic records.A number of numerical tests with irregular surfaces have validated the stability of this proposed method.The tests also show the simulated wavefields are more accurate than using the vacuum method,which is another commonly used method for dealing with an irregular surface.(2)A robust algorithm for first-arrival waveform inversion is developed for land seismic data by incorporating the improved “immersed boundary” method.The algorithm uses the intermediate data to solve the inconsistency problem between observed data and predicted data,as well as,avoid the cycle-skipping problem.Besides,the waveform inversion system is derived for using the land geophone data,which contains the particle vibration velocity.Both synthetic and real data applications demonstrate the high robustness of this proposed algorithm.(3)The surface-wave waveform inversion system is established for ambient noise data.A multi-component 3-D elastic full-waveform inversion system is developed to invert the surface waves,which are extracted from ambient noise data.The spectral element method(SEM)is used to simulate the elastic wavefield with an irregular free surface.In order to validate the above three research achievements,the first-arrival waveform inversion is applied to three land field seismic surveys.The inversion results show that the first-arrival waveform inversion produces velocity models with higher resolution and deeper effective update depth than the ray-based tomography method.These recovered velocity models could be used as an accurate shallow velocity model for subsequent seismic data processing.In addition,to verify the surface-wave waveform inversion system,the ambient noise data recorded near the San Jacinto fault in the United States was inverted.Inversion results show that the developed surface-wave waveform inversion system can construct a 3-D S-wave velocity structure with higher resolution and better waveform fitting than the 1-D dispersion curve inversion.In summary,the proposed methods including the forward modeling and waveform inversion with an irregular surface have been successfully applied in several real study areas.This demonstrates that the works in this thesis are effective experiments and explorations of seismic waveform inversion for land field data.