Full waveform inversion (FWI) is a strongly nonlinear optimization problem, which suffers from cycle skipping when the initial velocity model is not good enough or the seismic data lack low frequencies. Traveltime tomography is often used to invert the low‐wave number components of a velocity model, but traveltime inversion results show a low‐resolution model. To bridge the inversion wave number gap between the traveltime tomography and the FWI, a wave equation‐based local travetime inversion is proposed. In this case, the cross‐correlation algorithm is applied to calculate the traveltime differences between the localized synthetic and observed data. To mitigate traveltime crosstalk noise caused by multiple seismic events, a sliding Gaussian window is applied to transform the seismic data into local domain. In this way, the traveltime information of different seismic events can be simultaneously used to improve the inversion results. The numerical examples show that the combination of Local Traveltime Inversion (LTI) and FWI successfully mitigate the cycle skipping and produce satisfactory inversion results, even if the seismic data lack low frequencies and the initial velocity model is far from the true one.

中文翻译：

---

基于波动方程的局部行进时间反演

全波形反演（FWI）是一个强烈的非线性优化问题，当初始速度模型不够好或地震数据缺乏低频时，会遭受周期跳跃。行进时间层析成像通常用于反转速度模型的低波数分量，但行进时间反演结果显示出低分辨率模型。为了弥合行进时间层析成像和FWI之间的反演波数差距，提出了一种基于波动方程的局部旅行时间反演。在这种情况下，将使用互相关算法来计算局部合成数据和观测数据之间的传播时间差。为了减轻由多个地震事件引起的传播时间串扰噪声，应用滑动高斯窗将地震数据转换为局部域。通过这种方式，不同地震事件的传播时间信息可以同时用于改善反演结果。数值算例表明，即使地震数据缺乏低频且初始速度模型与真实速度模型相距甚远，局部行进时间反演（LTI）和FWI的组合仍能成功地缓解周期跳跃并产生令人满意的反演结果。