SEG Distinguished Lecture

Land seismic data is often dramatically noisier than marine data. Processed images and attributes from land data can be severely degraded by this noise. The acquisition effort needed to address the noise is often the biggest cost factor and the biggest uncertainty with land seismic data. 

The physics behind land seismic noise is fascinating. Much of the noise is caused by the very near surface, less than 30m, which often has ultra slow velocities. This land seismic noise is often complex and irregular — much more so than simple surface waves. Even simple scattering models are insufficient. We give a physical intuition of this noise generation through wave propagation movies and synthetic data examples with realistic models. Better understanding the physics can lead to better appreciation and customization of the acquisition and processing for the noise types.  

We propose the following conclusions for many areas: 

1. Micro-near surface effects are important
2. Surface distortion of signal is more complex than the statics, SCamps, and SC-Decon model and this can be strong
3. You probably have more signal in your data than you think
4. Statistical noise removal has limits
5. There is much value in coherent noise removal
6. There is limited benefit to dense inline receiver spacing
7. Aa better use of high channel counts may be an irregular layout
8. Noise character may limit CSI-reconstruction approaches
9. Changes in surface moisture content can produce strong 4D noise for CCUS monitoring
10. FWI has limited benefit in addressing the surface noise (but it has other good uses)
11. Specialized modeling of the noise in an area and testing methods to address it can be helpful.