论文部分内容阅读
给出一种稳健反演方法(也称迭代加权最小二乘IRLS),该方法对分离数据点极不敏感,因为它可以自动剔除异常的数据点,不需要人工仔细检查和控制数据质量。为了使用Tarantola-Valette的迭代广义离散反演方法进行稳健反演,将基于残差和信噪比的不同判据引入到协方差矩阵(作用相当于权重函数)中。在实际应用中,这个算法的作用相当于震源机制初步测定(PDFM),后者在海啸预警中用于快速估计强震震源参数。要反演的输入数据是长周期面波的频谱,输出的计算结果为地震矩张量,根据这些张量获得地震的震源几何形状和主应力轴。这种算法可以应用于任何非线性(迭代)反演计算中,解决污染数据集的分离数据点的问题。
A robust inversion method (also known as iterative weighted least squares IRLS) is presented, which is extremely insensitive to the separation of data points because it automatically rejects anomalous data points without the need for manual scrutiny and control of data quality. To perform robust inversion using Tarantola-Valette’s iterative generalized discrete inversion method, different criteria based on residuals and signal-to-noise ratios are introduced into the covariance matrix (acting as a weight function). In practice, this algorithm acts as a primary source mechanism measure (PDFM), which is used to quickly estimate source parameters of strong earthquakes during a tsunami warning. The input data to be inverted is the spectrum of long-period surface waves and the output is calculated as the moment tensor of the earthquakes, from which the source geometry and principal stress axes of the earthquakes are obtained. This algorithm can be applied to any non-linear (iterative) inversion calculation to solve the problem of separating data points from contaminated data sets.