Geology & Geophysics Editors' Highlights

Machine Learning for Magnetics

Classic interpretation of aeromagnetic anomaly maps involves several steps with limiting boundary conditions; a recent study develops convolutional networks largely bypassing these issues.

Source: Journal of Geophysical Research: Solid Earth


Interpretation of (aero)magnetic anomaly maps typically involves a series of steps after optimization of raw flight data. Nurindrawati and Sun [2020] explore the performance of machine learning methods for this kind of problem. They test various sets of convolutional neural networks (CNN) and train the two optimal CNN designs (one for declination and one for inclination) to predict declination and inclination directly from input total field maps. No reduction to pole or similar data pretreatment is required. Classic magnetic anomaly interpretation is rather often critically impacted by constraints on the depth and shape of the anomaly source body.

Importantly, the authors illustrate that their CNN approach is less influenced by source body shape, under the proviso the center of the source body is reasonably placed, for example as inferred from existing depth estimate methods. The CNN approach is anticipated to develop into a welcome addition to the existing tool kit for magnetic anomaly interpretation.

Citation: Nurindrawati, F., Sun, J. [2020]. Predicting magnetization directions using convolutional neural networks. Journal of Geophysical Research: Solid Earth, 125, e2020JB019675.

—Mark J. Dekkers, Associate Editor, JGR: Solid Earth

Text © 2020. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.