【作者】 谭毅培；

【导师】 陈棋福；

【作者基本信息】 中国地震局地球物理研究所 ， 固体地球物理， 2012， 博士

【摘要】 历史地震宏观烈度资料指史料中对地震破坏程度的记录,是研究历史地震最基本也是最重要的数据来源之一。有仪器记录地震波形的时间相对短暂,对历史地震震源参数的估算为地震活动性和地震危险性分析提供重要的基础资料。本研究在分析现有历史地震震源参数估计方法的优势、不足和应用范围基础上,利用贝叶斯决策论、模型选择和多模型统计推断的原理和算法,设计了一套估计历史地震震源参数的方法。本研究首先描述初步估计可能发震断层走向的贝叶斯判别方法。此方法依据贝叶斯最大后验概率原则,直接根据烈度数据点空间分布特征构建烈度判别方程,并画出方程代表的二次型判别曲线,根据判别曲线几何形态估计可能发震断层走向,无需先行确定震中位置或震级等其它震源参数,计算过程具有可重复性。方法经过设定走向的数值恢复测试,判别曲线走向中心值估计结果较好,参数估计具有一定统计无偏性。通过8个有烈度调查记录和波形反演震源机制解的地震实例检测,在其中7个震例中估计结果可以接受。本研究借鉴直接拟合烈度数据点和枚举震源参数的做法,设计了估计历历史地震震源参数的模型选择方法。该方法对震源参数所有可能的组合进行枚举,采用地震波场模拟计算转换的理论烈度值,利用模型选择方法评估各可能的震源参数组合模型与历史破坏记录推断的地震烈度数据点的拟合程度,对震源参数做出估计。该方法充分考虑到历史资料相对稀少对震源参数估计的影响,以多种震源参数估计结果和相应权重值来定量化表示估计结果的不确定性。通过对给定震中位置、震源深度和滑动角的Bootstrap数值恢复检测,表明该方法得出的震源参数估计结果具有统计一致性和一定的无偏性。经过2004年美国Parkfield6.0级地震实例测试,所得参数估计结果与波形反演结果基本一致。本研究将该套方法应用于1882年河北深县6级地震的震源参数估计,基于历史资料推断的地震烈度数据点,分析认为深县凹陷区的北西向、东西向和北东东走向断层对烈度数据点拟合程度接近,三个走向的断层都有可能为1882年河北深县6级地震的发震断层。根据模型选择结果和地质资料的综合分析,东西向的旧城北断层或何庄断层,及北东东走向的深西断层为发震构造的可能性较大。更多还原

【Abstract】 Macroseismic data of historical earthquakes refers to earthquake damage record in historical materials, which is the most basic and important data source in historical earthquakes study. Compared with only available instrumental parameters of earthquakes from modern seismic networks in several decades, it is extremely important to determine quantitatively source parameters of earthquakes occurred in long history to seismicity investigation and seismic hazard assessment. That is particularly true for counties such as China with more than2000years historical documents of damaged earthquakes.In this study, we briefly review the investigated methods in historical earthquakes study with their advantages and shortcomings, then we herein described a scheme to estimate historical earthquake source parameters based on basic principles and algorithms of Bayesian decision theory, model selection and multimodel inference.First we developed a method to preliminary estimate the strike of seismogenic fault using intensity data points and Bayesian decision theory. This method calculates the intensity discriminant expressions with the distribution of seismic intensity data based on Bayesian maximum posterior probability. The possible strike of seismogenic fault is analyzed according to the geometric shape of quadratic discriminant curves. The result of this method is repeatable without any other input of source parameters like epicenter or magnitude. This method passed the numerical tests for given strike values with random deviations, and the test of eight earthquakes with well inversion results of strike. Test results showed the method is unbiased statistically.Then a model selection method to quantitative estimate source parameters of historical earthquakes using macroseismic data and synthetic seismograms was described. This method enumerates all possible models with different source parameters (such as epicenter locations, focal depths and rakes), and obtains theoretic intensity distribution of each model from synthetic waveform modeling, then evaluates source models fitting degree for intensity data inferred from historical earthquake records to determinate its parameters. We used multiple model solutions and model weights to give quantitative uncertainty of source parameters caused by relatively scarce of historical information. This method was tested with bootstrap numerical cases for given sources with random deviations, and the well determined source parameters of the2004Parkfield M6.0earthquake. The test results showed that this method is robust statistically.Finally the scheme was applied to the1882Shenxian earthquake occurred in Hebei province, China. Source parameters estimation results shown that there were similar fitting degrees with different assumed strikes of seismogenic fault in NW, WE or NEE direction. Considered model selection results and geological results, the Shenxian earthquake is caused possibly by Jiuchengbei fault or Hezhuang fault, or Shenxi fault.更多还原