A new algorithm and computer code for seismic tomography in anisotropic inhomogeneous media was developed. The new tomographic approach is a generalization of classical isotropic seismic tomography which introduces spatially and directionally varying slowness. The velocity model was considered as a stack of homogeneous blocks in contact, with each block individually parameterized using background velocities of P- and S-waves and a set of 21 anisotropy parameters. The inverse problem was solved sequentially in five steps, using the velocity model from the previous step as the starting model for the subsequent step. These steps form a chain with increasing complexity: (1) isotropic homogeneous model; (2) isotropic velocity model with vertical velocity gradient; (3) 3-D inhomogeneous isotropic velocity model; (4) 3-D inhomogeneous model with uniform anisotropy; (5) 3-D inhomogeneous generally anisotropic model. The new algorithm was applied to real bulletin data of 18 seismic stations deployed in SW Iceland and operated favourably for the monitoring of local swarm-like seismicity. Next, the resolution, robustness and accuracy of the inversion were discussed using real and synthetic data. Real data inversion revealed a predominantly depth-dependent isotropic velocity background and additional general 3-D anisotropy. The parameterization of the medium was too flexible to allow for a reliable interpretation of the anisotropy inside the elementary blocks and a cluster analysis was applied to stabilize the inversion results. Three important clusters were identified as a result. The orientation of the anisotropy (fast and slow P-wave propagation directions) of two clusters coincided with the strike of the documented faults. The orientation of the anisotropy in the third cluster was interpreted as a consequence of the fluid dynamics around Kleifarvatn Lake. The P-wave anisotropy strength reached a value of \(\pm\) 5–8%.

开发了一种用于各向异性非均匀介质中地震层析成像的新算法和计算机代码。新的层析成像方法是经典各向同性地震层析成像的推广，它引入了空间和方向变化的慢度。速度模型被认为是一堆接触的均匀块，每个块都使用 P 波和 S 波的背景速度以及一组 21 个各向异性参数单独参数化。逆问题分五个步骤依次求解，使用上一步的速度模型作为后续步骤的起始模型。这些步骤形成一个复杂度越来越高的链： (1) 各向同性均匀模型；(2) 具有垂直速度梯度的各向同性速度模型；(3) 3-D非均匀各向同性速度模型；(4) 具有均匀各向异性的 3-D 非均匀模型；(5) 3-D 非均匀一般各向异性模型。新算法应用于部署在冰岛西南地区的18个地震台站的真实公告数据，并在当地群状地震活动监测中运行良好。接下来，使用真实和合成数据讨论了反演的分辨率、稳健性和准确性。实际数据反演揭示了主要依赖于深度的各向同性速度背景和额外的一般 3-D 各向异性。介质的参数化过于灵活，无法对基本块内的各向异性进行可靠解释，并应用了聚类分析来稳定反演结果。结果确定了三个重要的集群。两个星团的各向异性方向（快和慢 P 波传播方向）与记录的断层走向一致。第三个星团中各向异性的方向被解释为 Kleifarvatn 湖周围流体动力学的结果。P波各向异性强度达到了\(\pm\)  5–8%。