Statistical Analysis Of Seismic Pattern Based On Ionospheric Perturbation

More than five millions of earthquakes happen every yearworldwide.As one of the most powerful nature hazards,earthquakes,especially strong shocks,bring huge loss of life and property to the whole world.How to effectively recognize the seismic precursors is a global problem in the recent years.Whereas the anomaly of plasma?both ion and electron?in the ionosphere can be detected before and after the strong shocks,numerous researches on ionospheric precursors in relation to the seismic activities are carried out during the past decades.It is expected that,by taking advantage of the anomalies detection method based on ionospheric perturbation,the seismic pattern can be recognized,even the approximate location of the epicenter can be predicted before an earthquake happens.One of the most widely analyzed ionospheric datasets is measured by the French satellite DEMETER?Detection of Electro-Magnetic Emissions Transmitted from Earthquake Regions?,which contains more than sixty million records around the globe from June 2004 to December 2010.From the point of view of the Fourth Paradigm,this paper attempts to study the seismic pattern and find a recognizing method for epicenter-neighboring orbits during strong shocks.Detection points or small regions are used as research objects in numerous studies on seismic activities recognition.Due to the infrequency of strong shocks,the number of non-seismic data is far larger than the abnormal one,which results in the underfitting during the training of recognition model.Hence,it is desired to put forward more suitable approaches to make better use of original data.In contract to the existing approaches,epicenter-neighboring orbits and non-seismic orbits are defined as the analyzing objects,which avoids the underfitting performance caused by the unbalanced data distribution.Moreover,statistical analysis shows thatthe trajectory of the ionospheric parameter recorded along a half-orbit is determined by the magnitude and the latitude of the epicenter.Two criteria,?₃ and ?₄,which are based on skewness and kurtosis,are put forward to quantify the asymmetry and stability of ionospheric parameters along the half-orbit.t location-scale distribution is carried out to fit the quantitative indicators.Moreover,during the distribution fitting,the probable cause of two typical types of exception are analyzed and explained.The abnormal indicators,?₃ and ?₄ caused by the strong shocks locate at the tails of the probability density function?p.d.f.?according to the Gutenberg-Richter's law.In order to determine which magnitude the abnormal ?₃ and ?₄ are caused by,suitable pairs of thresholds are calculated by means of dynamic time warping?DTW?distance.Finally,three groups of experiments are applied to figure out the optimal region of strong shock analysis?the range of magnitudes,ionospheric parameters and the location of boundaries?.Based on the result of optimal region of strong shock analysis,a seismic classification-based method for recognizing epicenter-neighboring orbit is proposed.As data located along the edge of seismic regions can hardly be classified into abnormal dataset or non-seismic one,a sloppy classification may badly reduce the accuracy of model.These boundary data are regarded as the unlabeled data by means of safe semi-supervised support vector machines?S4VMs?with kernel combination,which helps obtain a better classification performance.Furthermore,error correcting output coding?ECOC?strategy is utilized to transform the recognizing problem into a series of binary classifications.Finally,three groups of comprehensive experiments are applied to validate the effectiveness of the method,which achieves overall true-positive rate?TPR?and false-positive rate?FPR?about 80% and 1.5%,respectively.To be mentioned,the best performance happens when the quantity of unlabeled orbits is twice as large as the number of labeled ones.