The 35th annual conference on Astronomical Data Analysis Software & Systems 2025 in Görlitz, Germany

The lithosphere–atmosphere–ionosphere coupling (LAIC) mechanism posits that pre-seismic electromagnetic disturbances may manifest as detectable anomalies in the ionospheric magnetic field. The European Space Agency’s SWARM constellation offers a unique platform for probing these phenomena through high-precision, multi-satellite magnetic field observations.
We introduce a comprehensive computational framework for identifying potential earthquake precursor signatures in SWARM magnetic field data, leveraging advanced statistical and machine learning methodologies. Multiple anomaly detection algorithms are implemented and systematically validated to detect electromagnetic perturbations that may precede seismic events within the Dobrovolsky preparation zone.
The framework adopts a multi-tiered analytical approach that integrates Statistical Process Control (SPC), Isolation Forest machine learning, Principal Component Analysis (PCA), and spectral analysis. It processes magnetic field residuals derived from the CHAOS geomagnetic field model, with stringent quality control guided by geomagnetic activity indices . The pipeline accommodates user-defined temporal windows for pre- and post-seismic evaluation, with spatial constraints determined by the empirical Dobrovolsky scaling $R=10^{0.43M}$.
The system efficiently handles multi-satellite magnetic field datasets from SWARM Alpha, Bravo, and Charlie through parallelized computation, enabling large-scale event analyses. Statistical validation demonstrates the framework’s capability to detect anomalous magnetic signatures while accounting for inherent geophysical variability. The Isolation Forest algorithm achieves robust multivariate anomaly detection with tunable contamination parameters, while PCA facilitates dimensionality reduction and interpretation of complex magnetic relationships.
This framework constitutes a important advancement in computational infrastructure for earthquake precursor studies, offering versatile tools for investigating lithosphere–atmosphere–ionosphere coupling processes. Its multi-method design permits rigorous evaluation of precursor hypotheses. Future developments will focus on (1) integrating formal statistical significance testing to quantify anomaly reliability and (2) implementing machine learning–based predictive models for near-real-time earthquake probability forecasting. These enhancements aim to transition the framework from a retrospective analysis tool to a predictive platform for operational seismic monitoring and early warning.