Multi-parameter seismic full-waveform inversion using an integral equation approach
PhD-Candidate Kui Xiang
Morten Jakobsen (UiB)
Geir Nævdal (IRIS), Kjersti Solberg Eikrem (IRIS)
About the project
Full waveform inversion (FWI) has emerged as the ultimate answer to the Earth imaging and resolution problem. By performing a numerical simulation of the full wavefield including multiply scattered and diffracted waves in addition to diving and refracted waves and iteratively updating the seismic model until the computed waveforms matches the observed waveforms, one can potentially obtain images of the earth’s interior of much higher quality and resolution than when inverting travel times and/or amplitude data only. However, FWI is a very costly method and the iterative inversion results tends to be sensitive to the starting model.
By using multi-parameter full waveform inversion methods within the context of dynamic reservoir characterization, more accurate images of fluid movements and pressure effects in a reservoir during production can be obtained. The work of Jakobsen et al. (2015) is based on single parameter (variable velocity) acoustic approximations that we would like to use for the sake of simplicity when processing seismic data. In order to make the workflow more realistic for applications, we shall extend the use of the T-matrix method to complex media such as elastic and anisotropic media and multi-parameter full waveform inversion. In any case, this project represents a natural continuation of the pioneering work of Jakobsen et al. (2015).
Although we focus on petroleum related applications, the methods developed in this project are also relevant for a related project within medical ultrasound imaging as well as other projects within seismic monitoring of CO2 sequestration and seismic characterization and monitoring of geothermal reservoirs.
Figure: Simultaneous inversion of density and compressibility. Left: True model. Middle: Initial model guess. Right: Inverted result.
The main goal of this project is to develop methods for multi-parameter seismic full-waveform inversion using integral equation formulations. Special attention will be given to the development of methods for reducing the computational cost and the sensitivity of the inversion results on the starting model. Both deterministic and Bayesian formulations of the non-linear inverse scattering problems will be used. A further aim is to investigate the use of rock physics model in this context.