Automatic traveltime inversion via sparse decomposition of seismic data
Automatic traveltime inversion via sparse decomposition of seismic data
复制标题
通过地震数据的稀疏分解自动走时反演
DOI：
10.1190/geo2017-0329.1
复制
发表时间：
2018
期刊：
影响因子：
3.3
通讯作者：
Wang Huazhong
中科院分区：
文献类型：
--
作者：
Feng Bo;Wang Huazhong
我们开发了一种自动走时反演（ATI）方法来根据反射地震数据估计宏观速度模型。首先，我们使用稀疏分解方法提取局部相干事件的运动学信息（即源/接收器射线参数、传播时间和源/接收器坐标）。然后我们评估了一种基于射线相交准则计算反射走时残差的新策略，消除了地震振幅对走时残差估计的影响。可以通过使用基于梯度的方法最小化走时残差函数来迭代更新速度模型。为了获得没有伪影的平滑梯度，我们首先使用全变分（TV）正则化方法估计函数梯度的高波数分量，然后从全梯度中减去它。由于反射走时残差计算和速度更新是完全自动化的过程，因此所提出的走时反演方法被称为ATI。我们通过 2D 合成和现场示例确定 ATI 不需要一个好的起始模型。此外，它既不需要低频地震数据，也不需要长偏移距采集。然而，所提出的走时残差计算策略仅对2D情况有效，这限制了其3D适用性。我们探索 3D 扩展的可能解决方案。
We have developed an automatic traveltime inversion (ATI) method to estimate the macrovelocity model from reflection seismic data. First, we extract the kinematic information (i.e., source/receiver ray parameters, traveltime, and source/receiver coordinates) of locally coherent events using a sparse-decomposition method. And then we evaluate a new strategy to calculate the reflection traveltime residual based on a ray-intersection criterion, eliminating the influence of seismic amplitude to the estimation of the traveltime residual. The velocity model can be updated iteratively by minimizing the traveltime residual functional with a gradient-based method. To obtain a smooth gradient free of artifacts, we first estimate the high-wavenumber components of the functional gradient with a total variation (TV) regularization method and then subtract it from the full gradient. Because the reflection traveltime residual calculation and velocity update are fully automated procedures, the proposed traveltime inversion method is referred to as ATI. We determine with 2D synthetic and field examples that ATI does not need a good starting model. Furthermore, it requires neither low-frequency seismic data nor long-offset acquisition. Nevertheless, the proposed traveltime residual calculation strategy is only valid for the 2D case, which limits its 3D applicability. We explore a possible solution for 3D extension.