论文部分内容阅读
多事件定位法联合解决了震群中地震事件震源参数(震源和发震时刻)和台站记录到的地震事件的走时校正。本文阐述了基于网格搜寻技术的新的多事件定位方法。这个算法(称为GMEL)是比较早的单事件定位算法(GSEL)的拓展并且是基于定位问题中的最大似然公式表示。通过对使用线性反演(Geiger)方法的地方使用网格搜寻和根部查找技术,使GMEL适应拾取误差的非高斯模型和高斯模型,而且可以处理像预先理想定位事件的地面真实位置水平这样的先验约束。我们描述了这个算法,并报告了我们取得的成果,即在对多事件定位时,拓展了蒙特卡罗技术来计算事件定位的置信区间,通常它被用在GSEL方法中。这个扩展将会产生置信区间,它们量化了对震相拾取和估计台站校正的不确定性两方面在定位误差上的影响,也就是校准(或模拟)误差。我们比较了GMEL法和4个其他多事件定位法:震源分解法(HDC)、双差法(DD)、累进(progressive)多事件定位法(PMEL)和联合震源确定法(JHD),给出了检验的初步结果。这个检验应用到了1999年伊兹米特/迪斯杰地震序列的两个毗邻震群上,Engdahl和Bergman(2001)已经用震源分解法进行过分析。我们正在独立地将其他方法应用到他们使用的数据上,力图提高我们对于这些方法间存在的相似和不同以及通常多事件定位方法效果的理解。
The multi-event localization method jointly solves the travel time correction of the earthquake events recorded by the stations in the earthquake swarm with the source parameters (source and seismogenic moment) of seismic events. This article describes a new multi-event location method based on grid search technology. This algorithm (called GMEL) is an extension of the earlier Single Event Location Algorithm (GSEL) and is based on the maximum likelihood formula in positioning problems. By using grid search and root search techniques wherever the Geiger method is used, GMEL adapts to non-Gaussian and Gaussian models of picked-up errors and can process first the true location level of the ground like the pre-ideal location event Constraints. We describe this algorithm and report on the results we have achieved by expanding the Monte-Carlo technique to calculate the confidence interval for event location when locating multiple events, which is usually used in the GSEL method. This extension will yield confidence intervals that quantify the effect of positioning error on both the phase pick-up and the uncertainty of the estimated station calibration, ie calibration (or simulation) errors. We compared the GMEL method with four other multi-event localization methods: Source Decomposition (HDC), Double Difference (DD), Progressive Multi-Event Positioning (PMEL) and Joint Source Determination (JHD) The initial results of the test. This test was applied to two adjacent earthquake swarms in the 1999 Izmit / Disaster sequence and Engdahl and Bergman (2001) have been over-analyzed by source decomposition. We are independently applying the other methods to the data they use in an effort to improve our understanding of the similarities and differences that exist between these methods and the effects of the usual multi-event location methods.