We have formulated an inversion approach for near-offset seismic reflection data based on full-waveform inversion (FWI) workflows, in which an element of standard seismic processing takes the place of each of the fundamental components of FWI. A motivation for using reflections in FWI is trying to update the model in deep zones in which the diving waves could not penetrate due to the large-offset limitation. The workflow is an iterative cycle of simulation, depth migration, and impedance inversion, and it is an outgrowth of an approach referred to as iterative modeling, migration, and inversion. Two-way wave operators (similar to those used in reverse time migration), which are part of the gradient calculation, are replaced with one-way wave operators; pseudo-Hessian preconditioning of the gradient is replaced with a deconvolution imaging condition; and well-log information to calibrate the update direction (i.e., model validation) replaces the standard line search (i.e., data validation). For land data applications, variations in the source wavelet are anticipated to be a large issue. To address this, we update the amplitude and phase of the simulated source wavelet during the inversion. The alternative workflow is referred to as log-validated waveform inversion (LVWI). The stabilization of LVWI is analyzed in a synthetic environment in which we examine the sensitivity to geologic complexity, well location, and log interval, each of which can impact the usefulness of local well-log information. Data from the Hussar experiment, a 2D land survey carried out in 2011 specifically to test waveform inversion procedures, are used to validate the approach. The Hussar data set is the framework for several experiments designed to analyze the effect of the wavelet update and the well-log calibration procedures. The results suggest that the implementation of wavelet updates and well calibration may provide an opportunity to mitigate cycle skipping and overcome potential local minima.

中文翻译：

---

具有小波相位和幅度更新的反射地震数据的测井验证波形反演

我们已经制定了一种基于全波形反演 (FWI) 工作流程的近偏移距地震反射数据反演方法，其中标准地震处理的元素取代了 FWI 的每个基本组成部分。在 FWI 中使用反射的一个动机是尝试在深区更新模型，在这些深区中，由于大偏移量限制，潜水波无法穿透。该工作流程是模拟、深​​度偏移和阻抗反演的迭代循环，它是一种称为迭代建模、偏移和反演的方法的产物。作为梯度计算的一部分的双向波算子（类似于逆时偏移中使用的那些）被单向波算子取代；梯度的伪 Hessian 预处理被反卷积成像条件取代；和测井信息校准更新方向（即模型验证）取代标准线搜索（即数据验证）。对于土地数据应用，源小波的变化预计将是一个大问题。为了解决这个问题，我们在反演过程中更新了模拟源小波的幅度和相位。替代工作流程称为对数验证波形反演 (LVWI)。在合成环境中分析 LVWI 的稳定性，在该环境中我们检查对地质复杂性、井位和测井间隔的敏感性，每一个都会影响本地测井信息的有用性。来自骠骑兵实验的数据，2011 年进行的 2D 土地调查专门用于测试波形反演程序，用于验证该方法。Hussar 数据集是多个实验的框架，旨在分析小波更新和测井校准程序的影响。结果表明，小波更新和井校准的实施可能提供一个机会来减轻周期跳跃和克服潜在的局部最小值。