Full text loading...
The laterally constrained inversion (LCI) method is widely recognized and applied for its robustness and reliability in processing TEM data. In the LCI method described in this study, the lateral constraints are incorporated into the inversion process using prior distributions within a Bayesian framework. The strengths of these lateral constraints, which control the smoothness of the model, are considered as hyperparameters in a Hierarchical Bayes algorithm. This allows for a data-driven estimation of the strengths of the lateral constraints based on the information contained in the observed data.
By incorporating Hierarchical Bayes in the LCI implementation, we achieve an adaptive strength of constraint that varies across the model space. This adaptability allows us to obtain a quasi-layered model while accurately resolving the lateral interfaces that are strongly supported by the available datasets. We conducted a comparative analysis using a synthetic example. We compared the inversion results obtained with adaptive constraints to those obtained with fixed constraints. The results include the median model and its corresponding uncertainty estimates, providing a comprehensive understanding of the model’s reliability.