**Study on Real-time Numerical Shake Prediction for Earthquake Early Warning**

Author:Wang Tianyun

Supervisor:Venus

Database:Doctor

Degree Year:2018

Download:78

Pages:105Size:7182K

Keyword:data assimilation，Ground Motion Prediction，Monte Carlo，Radiative transfer theory

Numerical shake prediction of shake map provides important information of the peak ground motion which indicates the potential disaster for Earthquake Early Warning(EEW)systems before the strong ground motions arrived the target region.Numerical shake prediction of shake map in real time can calculate the seismic ground motion of each region,and provides the more correct information for earthquake early warning system(EEW),avoid the potential disaster to soften the losses before real strong shakes arrived.The methods based on the estimated source parameters are calculated by the peak ground motions used the ground motion prediction equation(GMPE)of the empirical relationship with the observational records after first several seconds of P-wave detection.If source parameters are predicted error,traditional methods fail to predict the peak ground motion.The earthquake with large fault rupture leads to under-prediction of ground motions,and the magnitude saturation impacts on the predicting result.In 2015,Hoshiba investigated numerical shake prediction method based on radiated transfer theory and data assimilation technology for EEW system to solve this point.This method can predict the seismic wave strength distribution without estimating the source parameters based on sense observational stations.In most area of China,the distribution of seismic observational stations is not dense enough.Domestic research for numerical shake prediction is still an important task for EEW system.This paper focuses on some key technologies for the prediction and updating the shake map in real time.How to predict the shake map more rapidly and precisely whether in the dense or sparse seismic network situation is the objective in this paper.Traditional method is used to predict the initial shake map for a preliminary estimation when P-wave detected by one or more stations.Simulation of the real-time wave propagation is used to update the peak ground motion.The contents are summarized as follows:(1).The framework of numerical shake prediction and updating is suggested in this paper.The framework is divided into two steps.In the first step,initial shake-map is calculated by source parameters estimation based on GMPE empirical relationship using the first 3 seconds P-wave information.In the second step,updating method corrects the shake map by wave propagation model every 1 second using real-time records of seismic stations.(2).In this article,the source parameters prediction is suggested to be based on Bayesian formulation and Markov Chain Monte Carlo(MCMC)sampling method by first 3 seconds of P-wave records information,and provides subsequent ground motion prediction,denoted as initial prediction.It can predict the ground motions relatively robust,even in the sparse seismic networks.The source information can be estimated during the earthquake.Separating estimation of each source parameter will increase the risk of the estimation error.The probability model is based on Bayesian formulation(Liu and Yamada,2014)for modeling all source parameters simultaneously,and Markov Chain Monte Carlo(MCMC)sampling method to the source samples for the subsequent ground motion prediction.Shake map can be calculated in real time even lack of the observation information.(3).Seismic wave propagation is modeled by radiative transfer theory(RTT),energy wave field of shake map is modeled by data assimilation by given the status of the current energy distribution on the shake map.In 2015,Hoshiba applied this numerical shake prediction method,and used in 2011 Mw9.0 Tohoku earthquake as an example.The computational time of the numerical shake method is less wasted.But this method is not applicable to most area of China which network is not dense enough.Source parameters are important complementary used in initial prediction of shake map in this paper.And source location can be used to correct the seismic wave propagation velocity in 2-D space.(4).Both 3-D and 2-D space models are compared to study in this paper,and the shake map of real-time prediction in 3-D space model shows more precisely predicted results.From past studies,wave direction is assumed to propagate on the 2-D surface of the earth for computation efficiency.In fact,real seismic wave propagates in the 3-D sphere of the earth.The seismic wave propagation assumed in 2-D space leads to inaccurate wave estimation results.A 3-D space numerical shake pre-diction method is proposed in this paper,which simulates the wave propagation in 3-D space model by radiative transfer theory,and incorporate data assimilation technique of estimation to the wave energy distribution.The prediction result shows that shake map prediction based on 3-D space model can predict the real-time ground motion precisely,and over-prediction is alleviated by using 3-D space model.(5).By using seismic observational records of 2008 Ms8.0 Wenchuan earthquake in China,and the 2011 Mw9.0 Tohoku earthquake in Japan,the feasibility and applicability of numerical shake map prediction method in this paper was verified and analyzed.This paper made contributions from the following four aspects,and have obtained some achievements :(1).On the basis of numerical shake prediction method,developed the framework of numerical shake prediction and updating in this paper,sought to achieve a good balance of the timeliness and accuracy.(2).Bayesian formulation and Markov Chain Monte Carlo(MCMC)sampling method is used to calculate the initial prediction based on the GMPE empirical relationship formulas.(3).Add the weighted particles sampling algorithm in the updating method,corrected the prediction in real time.(4).A 3-D space numerical shake prediction method is proposed in this paper,which simulates the wave propagation with Monte Carlo method in 3-D space model to estimate the wave energy distribution,and compared with 2-D space model.