Pre-earthquake Anomaly Detection And Statistical Analysis Of Swarm Satellite Magnetic Field Data

The earthquakes not only endanger the production and development of society,but also endanger the safety of human life.The earthquake prediction is a worldwide problem that requires long-term continuous research due to the complexity of the earthquake mechanism and the long-term recurrence of great earthquakes.Therefore,the study of earthquake precursors has great significance.A great number of studies have shown that electromagnetic anomalies before the earthquake not only exist in the lithosphere,but also appear in the atmosphere and ionosphere through coupling.With the launch of electromagnetic satellites at home and abroad,the study of ionospheric disturbances before earthquakes using satellites has become an important method for earthquake precursor research.Swarm is a satellite constellation mission launched by ESA(European Space Agency)on November 22,2013.It mainly measures the geomagnetic field including the main magnetic field,the lithosphere magnetic field and the external source field with high accuracy.Earthquakes happen in the lithosphere.To analyze the anomalies in the ionosphere caused by the earthquakes happened in the lithosphere,it is necessary to remove the main magnetic field and the change of external field mainly affected by solar and geomagnetic activities.Therefore,in this paper,the main magnetic field that accounts for about 95% of the Earth's magnetic field is removed by subtracting the IGRF(International Geomagnetic Reference Field)model,the interference of the ionosphere on daytime is eliminated by choosing the data which local time is nighttime as well.Finally,the principal component analysis is used to separate the magnetic fields of Swarm-A and Swarm-C into two principal components.The influence of geomagnetic activity is eliminated by removing the first principal component which has high correlation with the geomagnetic indices.For the Ecuador earthquake,through long-term analysis of the second principal component of the magnetic field of all tracks at night in 2016,the mean and standard deviation of skewness and kurtosis are small,and the skewness and kurtosis of the undisturbed magnetic field are close to zero.Therefore,the S-K coefficient is constructed based on skewness and kurtosis to extract pre-earthquake anomalies.It is found that with the approach of the earthquake time,especially from 39 days before the earthquake,the S-K coefficient begin to increase rapidly.And the S-K coefficient values of 11 tracks exceed the threshold.Until a period of time after the earthquake,the S-K coefficient gradually recovers to about zero.In addition,the cumulative number of anomalous tracks show different growth rates before and after the earthquake,which shows accelerated growth before the earthquake and slow recovery after the earthquake.And the inflection point of the fitted curve is close to the time of the earthquake as well.At the same time,the cumulative number of anomalous tracks of random earthquakes indicates a linear increase.The magnitude of the anomalies decreases with the increase of the distance from the epicenter.These results show that the extracted anomalies may be related to the Ecuador earthquake.Based on the study of a single earthquake case,this paper performs pre-earthquake anomaly extraction and statistical analysis on earthquakes greater than magnitude 5 worldwide using the Swarm satellite magnetic field data from 2014 to 2017.First,anomaly extraction is performed on the second principal component of the magnetic field of more than 20,000 tracks in these 4 years.The ratio of the root mean square of the second principal component in the moving window to the entire track is calculated and a threshold is set to extract the anomalies,and the time and location of the anomalies are record at the same time.The time of anomalies and the distance between the epicenters and the anomalies are used to correlate them with the earthquakes,so as to study the statistical characteristic of the anomalies before the earthquakes.By analyzing the anomalies,it shows that among the 2228 earthquakes worldwide,the number of effective earthquakes is 1165,the earthquake detection rate is 52.29%,and the earthquake detection rate of random earthquakes is lower than the actual earthquakes.This result supports the credibility of the anomaly extraction results.According to statistics,the earthquake detection rate increases with the magnitude of the earthquake;and the shallow source earthquakes are more likely to produce ionospheric anomalies than deep source earthquakes.The number of anomalies decreases as the epicenter distance increases.The number of effective anomalies and effective earthquakes on the 90 th,34th and 32 th days before the earthquake are higher than usual.