Automatic classification of ionospheric anomalies potentially linked to earthquake occurrence using machine learning techniques

The ionosphere is one of the most important layers of the atmosphere, and for its electric properties is used in communication and navigation services. In addition to being influenced by geomagnetic and solar activity, in recent decades, it has been observed that the ionosphere may also exhibit variability due to effects caused by events of terrestrial origin, such as earthquakes. However, objectively identifying when a variation is an anomaly related to an earthquake remains a challenge.

This study presents a methodology based on machine learning to automatically detect the relationship between this type of irregularity and earthquakes. For this purpose, electron density (Ne) data recorded by the European Space Agency’s Swarm satellite constellation are used. Following the previously published NeAD anomaly detection algorithm, a combination of machine learning techniques is applied to group the detected anomalies according to their characteristics, to correct and automatically distinguishing the anomalies truly associated with the earthquake under study.

As a case study, the Mw 7.6 earthquake that occurred in Mexico on September 19, 2022, is presented. Five types of anomalies were distinguished, showing that duration and intensity are the most important factors for differentiating them.

The results suggest that one of the five anomaly groups can be associated exclusively with processes related to the main earthquake, while the other four groups are linked to other phenomena such as other minor earthquakes, tropical cyclones, or volcanic eruptions. 

This automated approach opens new possibilities for improving the classification of ionospheric anomalies and understanding how the lithosphere, atmosphere, and ionosphere interact with each other in the dynamics of our planet.