Lithologic prediction of strata based on hierarchical Bayesian modeling: an example of Carboniferous strata in the northern Junggar Basin

In geophysical exploration, accurately predicting lithology is essential.  The uncertainty in inverse solutions and variable physical properties across different strata often require manual lithology interpretation.  To overcome this, we developed a method integrating multiple geophysical inversion results for regional lithology prediction.  This method, applied to empirical data, enables lithological classification of Carboniferous reservoirs.  It utilizes custom spatial geological priors, discernible through well-logging and core geological principles.  Our approach, grounded in a hierarchical Bayesian model, includes data fitting and regularization components.  The data fitting component uses well-log data from proximate areas for training, and the regularization component incorporates established geological knowledge.  Our approach ensures that lithological classification is consistent with fundamental geological rules and aligns with inversion results.  We employed diverse inversion data types (electromagnetic, seismic, gravity, and magnetic) from the Carboniferous reservoirs in the northern Junggar Basin.  Our lithological predictions, validated against well-log data, offer a fresh perspective on integrated geophysical interpretation and diminish subjective bias in geological assessments.