An efficient inversion framework for audio-magnetotellurics with borehole constraints combining supervised descent method and Gaussian distribution modeling strategy

In audio-magnetotellurics (AMT) inversion, the resistivity model derived from data is crucial for understanding geological properties. Current AMT inversion methods like Gaussian-Newton and nonlinear conjugate gradient have limitations, including sensitivity to data errors and reliance on initial models, leading to non-uniqueness and slow convergence. To address these issues, we propose an AMT inversion framework incorporating borehole data and geological constraints. By leveraging borehole information and considering geological patterns, we develop three machine learning data construction methods that enhance the stability and speed of the inversion process. However, borehole data acquisition is costly and limited, and it represents geological properties discretely within a narrow range. Relying solely on it or unconstrained inversion can compromise accuracy. Our approach integrates borehole data into the SDM inversion, resolving data gaps and model variations. SDM results are then used as initial models for Gaussian-Newton inversion. Synthetic and field data examples demonstrate the efficiency and feasibility of our framework, showing rapid convergence and high-quality results. This approach accelerates AMT inversion by effectively utilizing borehole information, providing a practical solution for improving the process.