一种时间域全波形反演中的地震子波估计方法

全波形反演具有高精度成像能力,然而由理论走向实际应用还存在很多问题.全波形反演通过匹配波形来更新模型,数据的波形与地震子波有直接的关系.本文介绍了一种针对时间域波形反演的子波估计方法,并将其应用到全波形反演中.由于在频率域子波反演可表示为一个线性优化问题,因此本文先给定一个试探子波,在时间域通过有限差分法正演得到地震波场,并在频率域与观测数据比较分析,反演得到预测子波.此外,本文还简要地介绍了如何使用伴随状态法计算全波形反演的梯度.数值实验证明,本文方法反演得到地震子波与真实子波具有很好的吻合度,并在全波形反演中取得了不错的效果.相比不依赖于子波的方法,提前进行子波估计可提高反演效率.此外,本文方法适用于全波形反演中常使用的多尺度反演策略.

Seismic wavelet estimation for full-waveform inversion in time domain

Full-Waveform Inversion(FWI)has the great power to build high-resolution subsurface images.Wavelet estimation is one of the most important problems in FWI.In this paper,we introduce a wavelet estimation method for FWI,which can be organized as a linear optimization problem in frequency domain.We first use a tentative wavelet to obtain a trial wavefield in time domain,then invert the wavelet in frequency domain using the trial wavefield and the observed data.In addition,we also briefly give the introduction of finite difference method for waveform modeling and the derivation of gradient of FWI using the adjoint-state method.Numerical example has demonstrated the estimated wavelet with the proposed method is close to the true one.The applications of the estimated wavelet to FWI also show that the estimated wavelet can obtain similar inversion results as the true wavelet.Compared to the source independent method,the proposed method can accelerate the convergence.In addition,the proposed method is suitable for the multi-scale strategy,which is widely used in FWI especially when the initial model is poor.