Modeling and inversion of seismic data using multiple scattering, renormalization and homotopy methods (COMPLETED 14.05.2020)
PhD-Candidate Xingguo Huang
Morten Jakobsen (UiB)
Geir Nævdal (IRIS)
About the project
Seismic data are obtained when a pressure wave is sent into the Earth and the energy reflected at geological boundaries is recorded at the Earth’s surface. This principle is exactly the same as in medical ultrasound. When 3D seismic data are obtained at the same location at different times, the resulting data are referred to as 4D seismic data. Such data allow us to monitor how properties at a specific target area in the Earth’s subsurface changes with time.
In the last couple of decades we have witnessed an increased use of 4D seismic data. Traditionally, the result of successful interpretation of 4D seismic data has been a better understanding of the oil saturation in the reservoir, leading to identification of the water-flooded areas and pockets of remaining oil, and an improved understanding of compartmentalization of the reservoir. The interpretation of the reservoir properties and dimensions are crucial when making decisions for drilling new wells.
4D seismic data always contain a certain degree of uncertainty. The quantification of uncertainties in 4D seismic data is, however, not an easy task, since the seismic data are often the result of a complicated seismic processing workflow. This workflow may not be fully consistent for each recurrent seismic data gathering and processing.
One way to better assess the uncertainties involved in a seismic processing chain involves the use of so-called full waveform inversion methods. This method involves several steps. First, a model is created on a computer based on an assumption on what the subsurface geology is like. Next, so-called seismic forward modeling is performed where the computer models what the seismic data should look like when implementing the same survey as was used for the 4D seismic data of interest. The modeled seismic data is then compared quantitatively with the real seismic data, and updates are made iteratively to the computer-generated geological model until the differences between the modeled and real seismic data are minimized. This may then yield a good idea of what the subsurface geology is like.
Illustration of full waveform inversion. Top left: True model. Bottom left: Initial model guess. Right: Examples of how full waveform inversion improves the initial model guess so that it approaches the true model.
The first goal of this project is to develop seismic forward modeling methods for highly complex geology. Another goal of the project is to develop efficient methods for time-lapse full waveform inversion through a number of mathematical and physical methods. This may lead to an improved images of the subsurface and a better understanding of which mathematical and physical approaches are suitable for seismic inversion and imaging, in addition to yielding more accurate images of the subsurface.
Click here for a more technical description of Xingguo’s PhD-project.