Seismic groundroll prediction by interferometry and separation in curvelet domain

• Main Author: Yan, Jiupeng
• Language: English
• Published: University of British Columbia 2011
• Online Access: http://hdl.handle.net/2429/34685

Groundroll is a type of surface wave that propagates along the Earth’s surface. Groundroll usually has low frequency, low velocity and high amplitude. Due to its high amplitude, groundroll almost always dominates reflected body waves in land seismic data and covers important reflection information. Therefore, removing groundroll noise is a very important step before seismic imaging. The most common methods used in industry to remove groundroll are the Fourier domain filtering methods based on the different characteristics of groundroll and reflections, i.e. the low frequency and low velocity properties of groundroll. However, groundroll and reflection usually have large overlap in both physical and frequency domain. Also groundroll is spatially aliased at normal receiver intervals causing additional processing difficulties. Therefore, a good separation of groundroll by Fourier domain filtering method is challenging. In this thesis, we propose a data-driven workflow to remove groundroll. Our workflow is motivated both by SRME (Surface Related Multiple Elimination) method and a recently proposed interferometry method for the prediction of groundroll. It consists of a prediction step based on interferometry and a robust separation step that involves curvelet domain matched filtering and sparsity promotion. Tests of our workflow on synthetic data show clear removal of large amplitude groundroll and preservation of seismic reflection events. Test of our separation step on real data shows improvement over conventional Fourier domain filtering methods.