Pre-earthquake Anomaly Extraction And Evaluation Of Borehole Strain Observations

Earthquake disasters seriously threaten the safety of people's lives and property.Due to the small probability of earthquakes,the complexity of the seismogenic mechanism and the technical limitations of deep earth observations,earthquake prediction has become a worldwide problem that requires long-term exploration and research.After decades of observational technology development,China has accumulated a wealth of observational data in deformation,geophysics,and geochemistry disciplines,providing a data basis for the study of earthquake precursors.How to eliminate anomalies caused by environmental interference and other reasons from a large amount of observational data,extract the anomalies before earthquakes and realize the evaluation of earthquakes based on the anomalies,meet major social needs with profound scientific significance.An earthquake is a process of sudden rupture of an underground rock layer subjected to a concentrated stress exceeding the strength limit.Borehole strain records the changes in the crustal strain,which plays an important role in the study of earthquake precursors.However,there are still the following problems in the analysis of borehole strain observations before earthquakes:(1)In the observation of borehole strain,it is difficult to distinguish the strain caused by the earthquake due to the response of the strain response caused by the earth tide,air pressure and other environmental factors.(2)Extracting anomalies of borehole strain observation is mostly realized by an amplitude threshold over-limit method in the time domain,time-frequency domain.But the data amplitude is easily affected by accidental factors such as sudden interference or noise.(3)Most of the anomaly analysis use a single station method.Even if the anomaly analysis adopts the multi-station method,it is only confirmed by the multi-station results,and its essence is still a single-station analysis.Rather than joint analysis of data between multiple stations,the pre-earthquake related features between multiple stations have not been fully utilized.In view of the above problems,the paper solves the strain response generated by various environmental factors based on a state space model,by using the Kalman filter method to;using an anomaly analysis method based on probability distribution,extracts the ApNe anomaly of a single station;constructs a topological network of multiple stations,analyzes the time evolution of the network degree,and extracts the anomaly of the network degree.The main contents of the paper are as follows:(1)The strain observation is calibrated according to the observation principle of the four-component borehole strain and its self-consistency.The tide factor is calculated,and the influence of the strained solid tide on the borehole strain is analyzed.The influence of temperature and air pressure on the borehole strain observation is analyzed theoretically.The influence of the relationship between water level,rainfall and strain observation is discussed.(2)A method for solving environmental response of borehole strain observation based on Kalman filter is proposed.According to the state space model of borehole strain observation,the measurement equation is used to map the environmental impact from the measurement space to the state space.Then the Kalman filter method is used to continuously correct the estimation of the environmental state through the iterative error covariance matrix to solve the strain response of earth tide,air pressure and borehole water level in order to restore the crustal strain signal.The rationality of each environmental response is verified by comparing the tidal factor,the theoretical value of Hooke's law and the theoretical value of the pressure on the crustal load effect.(3)A method for extracting ApNe anomalies from borehole strain data of a single station is proposed.Research on the probability distribution characteristics of crustal strain shows that the background signal of crustal strain is Gaussian.According to the zero-ApNe feature of Gaussian signal,the high-order cumulant method is used to extract the ApNe anomaly of the crustal strain signal from a single station.The analysis of the Wenchuan and Lushan earthquakes shows that the accumulation of ApNe anomalies of each station accelerates in power-law form before the earthquakes,and the accumulation rate of anomalies has an inverse relationship with the epicenter distance of each station.Random time and random earthquake experiment also illustrate the correlation between ApNe anomalies before earthquakes.(4)An anomaly extraction method for pre-earthquake network degree of borehole strain data from multiple stations is proposed.A topological network of multi-station strain observation is established to avoid the contingency of anomaly from a single station.Multi-channel singular spectrum analysis is used to filter out the strain components in the observation data of multiple stations,and the Pearson correlation coefficient of the strain components is used to define the edge.The paper connects the edges according to the correlation coefficient threshold,and calculates the network degree and extracts the network degree anomaly based on the edges.The 6-station network degree anomaly analysis for 13 target earthquakes in Sichuan-Yunnan area shows that there was a power-law increase in network degree anomalies before the 10 earthquakes,and the anomaly accumulation rates are related to the target earthquake energy levels.(5)The abnormal b value can reflect the stress state in the source area.The ApNe anomalies of a single station before the earthquake is analyzed based on the b value evolution before and after the Lushan earthquake.For network degree anomalies,an earthquake detection model is established.It estimates the hit rate and false rate of earthquake detection with the receiver operating characteristic method.The analysis of 29 earthquake detections shows that the hit rates are much greater than the false rates,which indicates that network degree anomalies may contain pre-earthquake information.The efficiency of the earthquake detection model is verified after testing different parameters of the model.