Skip to main content

Coronavirus situation at the University of Eastern Finland

Vuoristolampi

Computational methods for seismic imaging and monitoring

Seismic imaging is a geophysical technique that aims at providing detailed maps of the Earth’s interior at diverse length scales with seismic waves generated by active or passive sources. This technique has numerous applications in the field of geophysics. It is widely used by the exploration industry to help guide the search for subsurface resources such as oil, gas, minerals, groundwater, and geothermal resources. In time-lapse mode, the technique is important for monitoring several time-varying processes in the subsurface, such as fluid movement and carbon dioxide (CO2) injection.

The most promising seismic imaging tool nowadays is Full Waveform Inversion or FWI, which utilizes optimization methods and computational wave propagation to reconstruct an Earth model that fits the entire content of recorded seismic data. Potentially, FWI can retrieve high-resolution subsurface images; however, the tool has its limitations. For example, the physics of seismic waves in the real Earth is complex, implying the realistic simulations of wave propagation are still computationally expensive. Moreover, most FWI methods are based on inexpensive but less accurate approximations such as constant-density acoustics, which begs the question, “What are the consequences on the reconstructed Earth model as a result”?

The doctoral dissertation of Kenneth Muhumuza, MSc, presents computational methods in seismic imaging and monitoring that overcome some limitations in the application of waveform inversion. An efficient modelling approximation known as the distorted Born method was employed to solve the problem of time-lapse seismic monitoring of CO2 injection using FWI. The objective here was to reconstruct the difference image of the subsurface between repeated active-source surveys acquired at different times to monitor the CO2 plume distribution. The reconstructed images showed that the method sufficiently retrieves the time-lapse velocity changes even in the cases of the relatively low signal-to-noise ratio.

A Bayesian approximation error approach was applied to recover from the consequences (and errors) associated with using acoustic Born FWI in a viscoelastic medium. This approach takes into account the modelling errors induced by these approximations and improved acoustic Born inversion for viscoelastic media, addressing elastic and viscous effects. The Bayesian framework is generally useful for estimating model uncertainties associated with FWI.

The dissertation also presents an emerging alternative to active source-seismic surveys known as ambient seismic noise imaging, which is a passive and low-cost approach. The technique uses ambient seismic noise from natural and anthropogenic sources for subsurface imaging and monitoring. Here, the continuous ambient noise data recorded by seismic stations in West Antarctica were used to infer the crustal structure of the Bransfield Strait. Findings agreed with those from previous research in the region.

The doctoral dissertation of Kenneth Muhumuza, MSc, entitled Computational methods for seismic imaging and monitoring will be examined at the Faculty of Science and Forestry. The opponent in the public examination will be Professor Børge Arntsen, Norwegian University of Science and Technology, Norway, and the custos will be Docent Timo Lähivaara, University of Eastern Finland. The public examination will be held in English in Kuopio on 28 October 2020 starting at 12 noon.

Photo available for download                           

Online event