Seismic imaging is a significant technology to provide the image of the subsurface in several fields such as hydrocarbon exploration/production and civil engineering. A fundamental problem in seismic imaging is that both the depth reflectivity and velocity distribution of the subsurface have to be predicted by only seismic events observed at the surface, and it still remains a challenging research topic. Joint migration inversion (JMI) is one of the seismic waveform imaging algorithms that were recently proposed. JMI is capable of simultaneously estimating velocity and reflectivity models of the subsurface by exploiting reflected waves including internal multiples. The seismic modeling algorithm in the JMI process is a method termed full wavefield modeling (FWMod), which is a one-way propagator-based reflection modeling algorithm, including higher-order scattering and transmission effects. In this thesis, two directions to improve the accuracy of seismic imaging based on JMI are discussed. On one hand, an extension of FWMod is proposed to correctly deal with not only reflected waves but refracted/diving waves via one-way propagators in the horizontal direction, and this method is extended to a new JMI algorithm. On the other hand, we assume that only reflected waves including internal multiples are utilized in the imaging based on JMI and present two novel methods for the inversion and pre-processing: 1) iterative reflectivity-constrained velocity estimation, 2) surface amplitude correction via learning from synthetic models for land seismic data. The reflectivity-constrained velocity estimation is employed to improve the accuracy of the estimated velocity by exploiting the estimated reflectivity in the JMI process. The surface amplitude correction process is introduced to mitigate the influence of the amplitude variations caused by source/receiver response sensitivities and the difference of the features between observed land seismic data and the simulated data by the used imaging scheme. The numerical and field data examples for both land and marine cases demonstrate that the proposed approach is capable of effectively estimating reflectivity and velocity model, even though the low frequency components of the observed data are absent.
|Award date||11 Sep 2018|
|Publication status||Published - 2018|