Robust inversion of 1D magnetotelluric data using Huber Loss function
Elfitra Desifatma, I Gede Putu Fadjar Soerya Djaja, Prihandhanu Mukti Pratomo, Enjang Jaenal Mustopa, and Wahyu Srigutomo*

Faculty of mathematics and natural sciences, Bandung Institute of Technology
Jl. Ganesha 10, Bandung 40132, Indonesia

Email of corresponding author:
*srigutomo[at]itb.ac.id

Abstract

In this study, a robust 1D inversion was performed on magnetotelluric (MT) data by utilizing the Huber loss function. In the MT method, impedance values (Z) obtained from the ratio between the electric field (E) and the magnetic field (H) that are perpendicular to each other are used to get subsurface information in the form of electrical resistivity variation with depth. Forward modeling of the 1D MT responses is carried out by recursively calculating Z values. Robust inversion is performed by using the Huber loss function as an objective function to minimize the difference between calculated data and observation data. Huber loss function combines squared loss and absolute loss to anticipate the presence of outliers in MT observation data. Robust inversion is performed on synthetic data with additional noise and real field data. The robust inversion scheme is successful in overcoming the existence of outliers and estimating the actual model parameters, both on synthetic data that has sufficiently large noise and on real field data.

Keywords: Magnetotelluric, Electromagnetic, Robust inversion, Huber Loss Function

Topic: COMPUTATIONAL SCIENCES