Flexible Coupling in Joint Inversions: A Bayesian Structure Decoupling Algorithm

Area of Science:

• Geophysics
• Inverse Problems
• Computational Science

Background:

• Simultaneous inversion of geophysical data constrains multiple physical properties.
• Determining spatial coupling between these properties is crucial for accurate subsurface modeling.
• Existing methods often assume either fully coupled or fully decoupled structures.

Purpose of the Study:

• To develop a novel algorithm for resolving the spatial coupling between geophysical properties.
• To enable the modeling of both common and separate structures within the same geophysical model.
• To assess the algorithm's performance against standard inversion techniques.

Main Methods:

• A Bayesian trans-dimensional adaptive parameterization forms the basis of the structure decoupling (SD) algorithm.
• The SD algorithm allows for a full spectrum of spatial coupling, from fully coupled to completely decoupled models.
• The approach was tested on 1-D geophysical inverse problems using synthetic and field data.

Main Results:

• The SD algorithm performed comparably to standard methods when structures were coupled.
• For decoupled structures, the SD algorithm successfully identified regions where physical properties did not share common interfaces.
• Application to field data demonstrated the algorithm's capability to decouple structures when common stratification was not supported.

Conclusions:

• The structure decoupling algorithm provides a flexible framework for geophysical inverse problems.
• It accurately models scenarios with varying degrees of spatial coupling between physical properties.
• This method enhances the interpretation of subsurface structures by differentiating shared and unique property distributions.