Exploration of Data Space Through Trans-Dimensional Sampling: A Case Study of 4D Seismics

Area of Science:

• Geophysics
• Subsurface Resource Monitoring
• Data Science

Background:

• Traditional subsurface resource monitoring often prioritizes model-space over data-space exploration.
• Understanding observational error structures is crucial for accurate subsurface analysis.
• 4D seismic data offers insights into subsurface variations but requires robust interpretation methods.

Purpose of the Study:

• To develop and present a novel data-space exploration methodology for 4D seismic data.
• To define and characterize structures within the covariance matrix of observational errors.
• To enable data-driven evaluation of 4D seismic survey quality (repeatability).

Main Methods:

• Bayesian inference utilizing trans-dimensional (trans-D) Markov chain Monte Carlo (McMC) sampling.
• Application of trans-D sampling to identify data-structures (partitions) in the covariance matrix.
• Utilizing laboratory-collected 4D seismic data to simulate various acquisition geometries and reservoir conditions.

Main Results:

• The trans-D sampling effectively defines data-driven data-space structures.
• The methodology successfully discriminates between different families of data-structures originating from various noise sources.
• Noise sources in 4D seismic data can be identified and differentiated.

Conclusions:

• The proposed methodology provides a data-driven approach to analyzing 4D seismic data structures.
• This method enhances the evaluation of 4D seismic survey repeatability.
• Integrating this data-space approach with model-space investigations can validate geophysical hypotheses for monitored geo-resources.