Bayesian Theory Based Earthquake Early Warning Uncertainty Analysis Method

As an effective means of disaster prevention,earthquake early warning has been paid more and more attention by researchers and government departments.In 2010,China launched the project of"national intensity rapid report and early warning project".At present,the national civil earthquake early warning system has been initially formed,and the earthquake early warning system for major projects is also in the research and experimental stage.For example,the earthquake early warning system for major projects such as high-speed railway,gas pipeline network and nuclear power project is also in the experimental stage.However,in order to obtain a higher timeliness,the earthquake early warning system attempts to use as little as possible trigger station information to estimate the parameters?magnitude,source location and ground motion parameters?of the earthquake that is occurring,which will inevitably bring greater uncertainty,which will bring confusion to the emergency disposal decision of major projects.For example,if the ground motion parameters are estimated too much,it will lead to unnecessary disposal and bring unnecessary economic loss;if the ground motion parameters are estimated too little,it will lead to failure to dispose in time and cause serious consequences.Therefore,how to reasonably estimate the uncertainty of the estimation results of earthquake early-warning parameters and provide a reasonable basis for decision-making is the key problem to be solved in earthquake early-warning research.In this paper,we use a large number of historical earthquake data to analyze the uncertainty in the methods of magnitude estimation,earthquake location and estimation of ground motion parameters,which are commonly used in the earthquake early warning system at home and abroad.The research contents and main achievements are as follows:1.This paper reproduces the method of estimating the magnitude of earthquake early warning based on the predominant period(?_p^max method),and establishes the statistical relationship between the predominant period and the magnitude by using 24673 vertical seismic records of 352 earthquakes recorded by Japan's K-Net strong motion observation network.The results show that when the magnitude is in the range of 2-6.5,the predominant period has a logarithmic linear relationship with the magnitude,and when the magnitude is greater than 6.5,it does not meet the logarithmic linear relationship.Therefore,this paper adopts the probability model of magnitude uncertainty of discrete magnitude,which makes the predominant period of each magnitude scale and its corresponding magnitude present lognormal distribution.The uncertainty analysis results of the combination of exponential distribution and uniform distribution of the prior distribution of magnitude are compared.The results show that the uncertainty of magnitude estimation results can be better reflected by the method of combining the uniform prior distribution of magnitude with the uncertainty of discrete magnitude.2.The real-time probabilistic positioning method of earthquake early warning based on the combination of equal time difference surface and UN triggered station information is reproduced,and the uncertainty of the positioning results of the method is analyzed by using the P-wave arrival information of 204 in network earthquakes and 22 out of network earthquakes recorded by K-Net strong earthquake observation network of Japan.The results show that the uncertainty of location results is lognormal distribution,and the uncertainty of location decreases with the increase of trigger stations;the residual of location of earthquakes in the network is small,the mean value of the residual is between 1-4km,and the residual of location of earthquakes outside the network is large.3.Considering the lack of large earthquake data in China,this paper collects large and small earthquake data and fault information of large earthquakes in China,Japan and New Zealand,establishes the attenuation law of seismic intensity of Chinese instruments based on the line source model,and analyzes the uncertainty of the attenuation law of seismic intensity of instruments.The results show that the residual distribution of the attenuation law of seismic intensity based on the line source model is normal,with the mean value of 0 and the standard deviation of 1.4.Based on the uncertainty analysis results of the above methods,the uncertainty probability model of earthquake early warning based on Bayesian formula is established,and the rationality of the uncertainty analysis model is analyzed by using the strong earthquake observation data.The results show that the uncertainty analysis model established in this paper can reasonably reflect the error propagation rule in the chain from the earthquake location and magnitude estimation to the estimation of ground motion parameters based on the attenuation rule,and can reasonably estimate the uncertainty in the earthquake early warning and provide the decision-making basis for the emergency disposal of major engineering earthquakes.