Small-scale lithospheric heterogeneity characterization using Bayesian inference and energy flux models
Small-scale lithospheric heterogeneity characterization using Bayesian inference and energy flux models
复制标题
使用贝叶斯推理和能量通量模型进行小尺度岩石圈非均质性表征
DOI：
10.1093/gji/ggab291
复制
发表时间：
2021
影响因子：
2.8
通讯作者：
González Álvarez I
中科院分区：
文献类型：
--
作者：
González Álvarez I
不同学科的观察表明，我们的星球在多个尺度长度上都具有高度异质性。尽管如此，许多地震地球模型往往不包括任何小规模的异质性或横向速度变化，这可能会影响基于这些同质模型的测量和预测。在这项研究中，我们根据固有的扩散和散射质量因子以及与特征尺度长度（a）和RMS分数速度波动（ε）相关的自相关函数来描述岩石圈小尺度各向同性非均质结构。为了获得这种表征，我们将单层和多层能量通量模型与新的贝叶斯推理算法相结合。我们的综合测试表明，该技术可以成功检索 1 层或 2 层模型的输入参数值，并且我们的贝叶斯算法可以解决数据是否可以通过一组参数拟合或需要一系列模型来拟合，即使对于非常复杂的后验概率分布也是如此。我们将该技术应用于澳大利亚的三个地震台阵：爱丽丝泉台阵 (ASAR)、瓦拉蒙加台阵 (WRA) 和皮尔巴拉地震台阵 (PSAR)。我们的单层模型结果表明，ASAR 的固有衰减和扩散衰减最强，而 ASAR 和 WRA 的散射和总衰减也同样强。 PSAR 的所有品质因数均高于其他两个阵列，这意味着该阵列下方的结构比 ASAR 或 WRA 的衰减和异质性更小。多层模型结果表明，对于所有阵列来说，地壳比岩石圈地幔更加不均匀。这些阵列的地壳相关长度和 RMS 速度波动范围分别为 ∼0.2 至 1.5 km 和 ∼2.3 至 3.9%。上地幔的参数值并不唯一，低参数值（a< 2 km 和 ε < ∼2.5%）的组合与具有高相关长度和速度变化的参数值（分别为 a> 5 km 和 ε > ∼2.5%）的可能性一样。我们将 ASAR 和 WRA 下方衰减和异质性结构的相似性归因于它们位于元古代北澳大利亚克拉通，而不是位于太古宙西澳大利亚克拉通的 PSAR。 ASAR和WRA下方小尺度结构的差异可以归因于同一克拉通这两个区域的不同构造历史。总的来说，我们的结果强调了能量通量模型和贝叶斯推理算法相结合对于未来散射和小规模异质性研究的适用性，因为我们的方法使我们能够获得和比较不同的质量因子，同时还为我们提供了有关确定散射参数时的权衡和不确定性的详细信息。
Observations from different disciplines have shown that our planet is highly heterogeneous at multiple scale lengths. Still, many seismological Earth models tend not to include any small-scale heterogeneity or lateral velocity variations, which can affect measurements and predictions based on these homogeneous models. In this study, we describe the lithospheric small-scale isotropic heterogeneity structure in terms of the intrinsic, diffusion and scattering quality factors, as well as an autocorrelation function, associated with a characteristic scale length (a) and RMS fractional velocity fluctuations (ε). To obtain this characterization, we combined a single-layer and a multilayer energy flux models with a new Bayesian inference algorithm. Our synthetic tests show that this technique can successfully retrieve the input parameter values for 1- or 2-layer models and that our Bayesian algorithm can resolve whether the data can be fitted by a single set of parameters or a range of models is required instead, even for very complex posterior probability distributions. We applied this technique to three seismic arrays in Australia: Alice Springs array (ASAR), Warramunga Array (WRA) and Pilbara Seismic Array (PSAR). Our single-layer model results suggest intrinsic and diffusion attenuation are strongest for ASAR, while scattering and total attenuation are similarly strong for ASAR and WRA. All quality factors take higher values for PSAR than for the other two arrays, implying that the structure beneath this array is less attenuating and heterogeneous than for ASAR or WRA. The multilayer model results show the crust is more heterogeneous than the lithospheric mantle for all arrays. Crustal correlation lengths and RMS velocity fluctuations for these arrays range from ∼0.2 to 1.5 km and ∼2.3 to 3.9 per cent, respectively. Parameter values for the upper mantle are not unique, with combinations of low values of the parameters (a< 2 km and ε < ∼2.5 per cent) being as likely as those with high correlation length and velocity variations (a> 5 km and ε > ∼2.5 per cent, respectively). We attribute the similarities in the attenuation and heterogeneity structure beneath ASAR and WRA to their location on the proterozoic North Australian Craton, as opposed to PSAR, which lies on the archaean West Australian Craton. Differences in the small-scale structure beneath ASAR and WRA can be ascribed to the different tectonic histories of these two regions of the same craton. Overall, our results highlight the suitability of the combination of an energy flux model and a Bayesian inference algorithm for future scattering and small-scale heterogeneity studies, since our approach allows us to obtain and compare the different quality factors, while also giving us detailed information about the trade-offs and uncertainties in the determination of the scattering parameters.