Imaging of small-scale heterogeneities with high-frequency seismograms and quantitative estimation of their physical properties

利用地震台阵数据，我们开发了一种新的方法来识别相干散射相位在高频尾波的峰值在频谱图的时间-频率域，并将它们成像为小尺度的非均匀介质。根据自回归模型的新方案，进行了背景介质的散射和/或衰减效应的校正以及频率-波数分析。该方法客观地选择了分析参数，提高了对各散射相强度和到达方向的分辨力。此外，我们利用三分量数据的优势，在第一时间获得每个散射模式的信息（即，P-to-P或P-to-S散射）。其结果是，我们成功地包括以下新的信息在小规模的异质性成像：提高空间分辨率，特别是在深度，与以前的研究相比，频率依赖性，和散射模式。将这种新的成像方法应用于日本仙台附近长町-利府断层周围地区的阵列数据，不仅得到了散射系数图，而且得到了它们与地震活动的关系。这些结果作为两篇论文发表在Journal of Geophysical Research上，我们进一步根据频谱图在时频域中的以下特征对成像的散射异常进行分类：（1）面积或扩展;（2）每个待成像峰的纵横比。这两个参数分别对应于许多小的子散射体之间的多次散射的程度和每个散射体相对于周围介质的内部速度。利用散射理论和数值模拟，我们将谱图中的观测参数与非均匀性的物理性质，如尺寸，速度或阻抗异常，形状或纵横比，以及聚类程度联系起来。

Using seismic array data, we develop a new method to identify coherent scattered phases in high-frequency coda as peaks in spectrogram of the time-frequency domain and to image them into small-scale heterogeneities in a medium. The correction of scattering and/or attenuation effects of the background medium and the frequency-wavenumber analysis are conducted, based on a new scheme with autoregressive models. This approach chooses parameters in analysis objectively and enhances the resolution in intensity and arrival direction of each scattering phase. In addition, we utilize the advantage of three-component data in the first time by obtaining the information of each scattering mode (i.e., P-to-P or P-to-S scattering in this case) with polarization analysis. As a result, we succeed to include the following new information in the imaging of small-scale heterogeneities: improved spatial resolution, particularly in depth, compared with previous studies, frequency dependency, and scattering mode. Applying the new imaging method to the array data in the area around the Nagamachi-Rifu fault near Sendai, Japan, we obtain not only the maps of scattering coefficients but also their relations with seismic activities. These results are in press as two papers in Journal of Geophysical Research.We further classify the imaged scattering anomalies based on the following characteristics of spectrograms in the time-frequency domain: (1) area or spreading and (2) aspect ratio of each peak to be imaged. These two parameters are shown to correspond to the degree of multiple scattering among many small sub-scatterers and the speed inside of each scatterer relative to the surround medium, respectively. Using scattering theory and numerical simulations, we relate the observed parameters in spectrogram with physical properties of heterogeneities, such as size, velocity or impedance anomaly, shape or aspect ratio, and the degree of clustering.