mahalanobis distance method for enhancing magnetotelluric data processing: a case study from the rochechouart impact structure, france

One of the primary challenges in magnetotelluric data processing is the presence of noise and outliers (anomalous values). these disturbances often come from human-made sources such as power lines, electronic devices, and nearby infrastructure. they can significantly affect the results related to apparent resistivity and phase, leading to unreliable models of subsurface electrical resistivity.to identify and remove these outlier and noisy components, the mahalanobis distance method is proposed as an effective solution. this approach — applied here in a four-dimensional feature space comprising the real and imaginary parts of two components of the impedance transfer function — calculates the distance of each data point from the dataset, meanwhile accounting for variances, covariances, and correlations between variables, thereby enabling the detection of anomalous points more accurately than simpler (2d) approaches.in this study, to identify outliers, we applied the mahalanobis distance method to real data from a single mt station located at the chassenon forage site within the rochechouart impact structure, france. the results demonstrate that this approach not only improves the accuracy of subsequent analyses and enables the extraction of more precise information from subsurface structures, but also reduces processing time by efficiently eliminating contaminated windows before final impedance estimation. one of the primary challenges in magnetotelluric data processing is the presence of noise and outliers (anomalous values). these disturbances often come from human-made sources such as power lines, electronic devices, and nearby infrastructure. they can significantly affect the results related to apparent resistivity and phase, leading to unreliable models of subsurface electrical resistivity.to identify and remove these outlier and noisy components, the mahalanobis distance method is proposed as an effective solution. this approach — applied here in a four-dimensional feature space comprising the real and imaginary parts of two components of the impedance transfer function — calculates the distance of each data point from the dataset, meanwhile accounting for variances, covariances, and correlations between variables, thereby enabling the detection of anomalous points more accurately than simpler (2d) approaches.in this study, to identify outliers, we applied the mahalanobis distance method to real data from a single mt station located at the chassenon forage site within the rochechouart impact structure, france. the results demonstrate that this approach not only improves the accuracy of subsequent analyses and enables the extraction of more precise information from subsurface structures, but also reduces processing time by efficiently eliminating contaminated windows before final impedance estimation.