The
First goal of the series:
To show that my existing non-linear inversion / coherent noise
removal works well, even
though major improvements are in the works.
The second: To show how
crucial seismic inversion is to effective interpretation, and to point out the
serious mapping errors that can occur because of tuning and noise.
The third: To
illustrate how non-linear (stochastic) logic opens one's eyes to the
serious problems that exist (such as tuning and coherent noise), and to
the understanding of how problems vary widely between prospects.
The fourth: To explain how
application of linear filtering has probably distorted the
data, making true inversion more difficult
Many old leases with missed reservoirs exist, where the data should be re-processed and new interpretations made. This includes much recent 3D shooting done on land and shallow water. Often this means going back to the original, pure data.
Probably a bunch. Is reprocessing easy? No it is not, and it takes a deep knowledge of the reflection process to make it worth while.
By going back to the raw data coherent noise can be evaluated and removed. Effective non-linear inversion can then be applied. The best way is to start at the beginning.
Everyone who depends on seismic should know about coherent noise, seismic tuning and other things that lead to errors in event picking, so plan to spend a little time with me.
On the far left is an example of seismic noise that is so strong it completely hides most of what we need to see. Noise almost always exists, especially on land and shallow water prospects, and it's often so serious that existing interpretations are not valid.
To its right are the results after non-linear refraction removal. So - just because older tries failed, don't give up too soon. Economics have changed, and we know how to do things better now, especially in noise removal and non-linear seismic inversion.
Even on clean data, what you see on a section is a complex mix of "tuned" primary events. This algebraic mixing can fool AVO logic and, at the minimum, create misleading dips. In fact tuning is the hidden enemy of seismic resolution, requiring effective seismic inversion to clarify stratigraphy and structure.
IN TODAY'S OVER-AUTOMATED SEISMIC WORLD "event picking" errors get passed to the mapping logic with little human control. Once there, they (the errors) are safe, never to be found. So my theme is "go back to the basics then leap forward". We are no longer looking for major anticlines, and resolution is worth a lot of human effort. In this case it involves analysis and software customizing.