中国科学院地质与地球物理研究所刘前程研究员团队，提出了利用尾波同时反演地震散射和吸收衰减的新方法。相关论文于2024年11月12日发表于国际顶尖学术期刊《中国科学：地球科学》杂志上。

据介绍，散射和吸收衰减的同时反演，对于研究地下的小尺度非均质性和非弹性性质具有重要意义。然而，将其应用于复杂的地球物理问题受到现有方法模拟和反演的昂贵计算需求的限制。在多参数反演过程中，散射系数和吸收系数之间的耦合效应导致参数串扰伪影，显著增加了反演过程的非线性。

本研究提出了一种快速有效的散射与吸收衰减同步反演方法。研究人员采用基于频域差分方程的有限元方法，以模拟尾波能量的传播。研究采用截断Gathems-Newton技术同时估计散射和吸收系数，以解耦反演过程中的两个衰减参数。然而，如果由于模拟和测量能量之间的相位不匹配而导致能量频率不合适，则反演可能达到局部最小值。

为了缓解这个问题，研究提供了一个频率选择标准，该标准考虑了能量密度频谱向较低频率的实质性频谱移位。基于合成和实验数据的数值算例表明，该方法无需精确的初始衰减模型，即可显著降低计算复杂度并抑制双参数串扰。

附：英文原文

Title: Simultaneous inversion of seismic scattering and absorption attenuation using coda energies

Author: Jia WEI, Qiancheng LIU, Ling CHEN, Liang ZHAO

Issue&Volume: 2024/11/12

Abstract: Simultaneous inversion of scattering and absorption attenuation is of great significance for investigating small-scale inhomogeneities and inelastic properties of the subsurface. However, applying this to complex geophysical issues is constrained by the costly computational requirements for simulations and inversions using existing methods. The coupling effects between the scattering and absorption coefficients cause parameter crosstalk artifacts in multi-parameter inversion, significantly increasing the nonlinearity of the inverse processes. This paper proposes a robust and effective simultaneous inversion method for scattering and absorption attenuation. The propagation of coda energy is modeled using the finite-element method based on the frequency-domain diffusion equation. We employ the truncated Gauss-Newton technique for the simultaneous estimation of scattering and absorption coefficients to decouple the two attenuation parameters during the inversion procedure. Nevertheless, the inversion may reach a local minimum if the energy frequency is inappropriate due to a phase mismatch between the simulated and measured energies. To mitigate this issue, we provide a frequency selection criterion that considers the substantial spectral shift of the energy density spectrum toward lower frequencies. Numerical examples using synthetic and experimental data indicate that our method can significantly reduce computational complexity and suppress two-parameter crosstalk without requiring a precise initial attenuation model.

DOI: 10.1007/s11430-023-1436-8

Source: https://www.sciengine.com/10.1007/s11430-023-1436-8