Online Only

NOTE: You Must Be Logged In to Register.

Speaker: Stephen Chiu, In-Depth Geophysical

Seismic data bandwidth controls the resolution of imaging subsurface geological structures. Unfortunately, the dispersive earth media typically reduce recorded seismic data to be band-limited in which both low and high frequencies tend to have poorer signal-to-noise ratio. Recent development of full waveform inversion often requires low-frequency components for converging to a reliable geological model. Thus, it is critical to recover low-frequency events for FWI applications. However, extrapolating reliable low-frequency signal is still a challenging issue.

We present a practical and effective method to extrapolate low frequencies on band-limited data. The basis of the extrapolation assumes that missing low-frequency signal is analogous to known higher-frequency signal. This method first derives a multi-channel prediction filter from known higher frequencies and uses the prediction filter to extrapolate the low frequencies. It operates in either Radon or XT (spatial and time) domain. Synthetic datasets and a field dataset have demonstrated that this method is robust in extrapolating low-frequency events. Using one of the simple synthetic data, we also compare the proposed method with a machine learning algorithm, a recurrent neural network (RNN) designed to use higher frequencies to predict lower frequencies. For the field data example, we will show how to integrate both low and high frequency extrapolations in achieving much high-resolution images. The results of the proposed method offer a simple and effective technique to extrapolate reliable low-frequency events that closely resemble the data structure of the higher-frequency events. This gives us the confidence that the extrapolated low-frequency events are real and not artificially created.

Speaker Biography: Stephen Chiu, In-Depth Geophysical
Stephen Chiu received a BSc in Geophysics from University of Saskatchewan in 1980; MSc in Geophysics from University of Alberta in 1982; and a PhD in Geophysics from University of Alberta in 1985. He worked for several seismic service providers in Calgary as a research geophysicist from 1985 to 1997. He joined ConocoPhillips in 1997 and spent almost 19 years working from a senior research position to the position of a principal geophysicist.  From 2016 to present, he serves as a principal geophysicist at In-depth Geophysical, Inc. His research experiences spans all facets of geophysical software development and applications, as well as data processing. His current research interests include developments in advanced signal processing and machine learning algorithms. He is an active member of the SEG and held several patents and published over 50 publications.
 

**Access information will be sent to all registrants after registration closes.

Price List: