[M-IS09] Interdisciplinary studies on pre-earthquake processes

[MIS09-12] Study on Seismic Anomaly Extraction of Swarm Satellite Magnetic Field Data Based on Complex Non-negative Matrix Factorization

*Donghua Zhang¹, Kaiguang Zhu¹, Yiqun Zhang¹, Baiyi Yang¹, Ting Wang¹, Wenqi Chen¹, Pu Wang¹, Yuqi Cheng¹ (1.Jilin University)

Keywords:Complex non-negative matrix factorization, precursors, Swarm satellite magnetic field data, 2016 Ecuador earthquake

To obtain earthquake-affected signal from the satellite observed data in the ionosphere, it’s necessary to avoid the influence of global disturbances in complex space environments of the ionosphere, such as geomagnetic disturbances. However, earthquake-affected signal and global disturbances possibly overlap frequency bands, which makes it difficult to separate them using frequency domain decomposition methods. Therefore, we use the magnetic field data of Swarm Alpha during the M7.8 Ecuador earthquake that occurred on April 16, 2016. Based on the spatial differences between global disturbances and local seismic anomalies, matrix factorization methods are used to obtain the local features that are possibly related to earthquakes from the observation data to avoid the impact of global disturbances. Compared to the traditional solution that only uses observation data from low geomagnetic level periods, this method improves the utilization of data. Meanwhile, as seismic information exists in both the amplitude and phase of the magnetic field, complex nonnegative matrix factorization in matrix factorization methods is selected to separate global disturbances and earthquake-affected signal in complex domain magnetic field data improving the utilization of magnetic field information, thereby enhancing the accuracy of identifying earthquake-affected signal. This method decomposes the observed data into three components, extracts anomalies from the three components, and accumulates the anomalies over time. It is found that the cumulative curve of one component has a strong correlation with the cumulative Benioff strain curve, showing a trend of accelerated growth before the earthquake and slow growth after the earthquake, indicating that this component is likely to be an earthquake-related signal.