Anisotropy identification in magnetotelluric data using deep learning methods

Anisotropy is a prevalent phenomenon in geophysical investigations, serving a crucial function in geological interpretation and geophysical inversion methodologies. Given its significance in accurately characterizing subsurface structures, there has been substantial scholarly focus on anisotropy. In the present study, we delineate an approach leveraging deep learning techniques to discern anisotropic structures from magnetotelluric responses. Parallel to conventional data-driven deep learning inversion methodologies, we curated a sample set of twenty million instances for neural network training. Subsequent sample evaluations were undertaken to validate the network's generalizability, robustness, and reliability.