Estimation of Horizontal Multilayer Soil Parameters Using Bayesian Inference

The inversion of earth resistivity structure is of great importance for the calculation of low-frequency electromagnetic interference on ground-based infrastructure. This article presents a Bayesian regression approach for the parameter estimation of horizontal multilayer soils. This supervised learning algorithm can provide more comprehensive statistical properties of soil parameters compared with classical optimization methods. The posterior probability distribution of the soil parameters is inferred by combining their prior knowledge with the measured apparent resistivity from Wenner's method. It allows for statistically quantifying the influence of measurement errors and shielding effects. Furthermore, the optimal number of layers can be distinguished using information criterion, considering both the goodness-of-fit and model complexity. Several multilayer Earth structure cases are used to illustrate the performance of the Bayesian inference method.