Exploration of the 2021 Mw 7.3 Maduo Earthquake by Fusing the Electron Density and Magnetic Field Data of Swarm Satellites

Earthquake is a complex and multivariate problem. Using a single parameter to extract anomalies is difficult to completely and truly reflect the preparation activity before earthquakes. In this article, we develop a fusion anomaly extraction method based on principal component analysis (PCA) and nonnegative tensor decomposition (NTD). It extracts anomalies by combining features of different parameters, which can obtain earthquake-related signals from the dataset and reveal some weak anomalies hidden in individual parameters. By PCA-NTD, we fused the electron density and magnetic field data from the Swarm satellites to explore the possible precursors of the 2021 Mw 7.3 Maduo earthquake and compared the results with those of the single-parameter analyses. The cumulative value of fusion anomalies indicates two acceleration stages before the mainshock: from -51 to -24 days, following a sigmoid trend, and from -21 to the earthquake occurrence, following a power-law behavior. The second acceleration is more pronounced than the first one, and its critical point occurs near the date of the Maduo earthquake. Spatially, these anomalies are located around important fault zones (Altun fault, Jiali fault, and Red River fault) and the epicenter region, which likely reflect the northward stress and eastward stress experienced in the seismogenic area before the mainshock. As the earthquake approaches, the anomalies become more concentrated and closer to the impending epicenter. Furthermore, the ionospheric anomalies correspond well with the anomalous phenomena of lithospheric activity and atmospheric thermal radiation, which support a multichannel lithosphere-atmosphere-ionosphere coupling (LAIC).