波动方程初至双差走时层析反演

对于有限频带的地震信号,地震波绝对走时受到震源子波的波形特征、走时测量准则等多种因素的影响,提高走时测量的精度有助于增加反演结果的可靠性。为降低绝对走时测量误差对反演的影响,采用双差走时代替绝对走时,并将其与Rytov近似导出的相位延迟敏感度核函数相结合,构造有限频带地震波的双差走时敏感度核函数。针对初至双差走时层析反演,提出了一种基于隐式矩阵向量积的高斯牛顿迭代算法,该方法无需显式计算、存储核函数及Hessian矩阵。相较于传统的散射积分方法(需要显式计算并存储核函数),波动方程初至双差走时层析反演方法仅利用了波动方程Born模拟及逆时偏移算法,因此降低了对计算机内存和外部存储的需求,可进行高效的大规模计算。高斯扰动模型数值模拟实验表明,该方法可以消除初至走时测量误差(震源子波未知)对反演的影响;Overthrust模型数据和二维起伏地表实际数据测试结果表明,采用该方法反演得到的高精度的近地表速度模型与实际地层较为吻合。

Wave-equation double difference first arrival tomography

For seismic signals with a limited frequency band,the seismic absolute travel time is affected by many factors such as the waveform characteristics of the source wavelet and the measurement criterion of the travel time.Increasing the accuracy of travel time estimation is essential for improving the quality of the inversion result.To reduce the influence of measurement errors of the absolute travel time on the inversion,the latter was replaced with the double difference travel time.The sensitivity kernel of the phase delay derived by the Rytov approximation was used to construct the double difference travel time sensitivity kernel of seismic waves with limited frequency.The new kernel not only inherited the advantages of the Rytov approximation but also reduced the inversion uncertainty.To solve the associated travel time inversion problem,an efficient Gauss-Newton algorithm relying on the implicit matrix-vector product method was proposed,which does not explicitly calculate the Hessian matrix.In fact,the conventional scattering-integral method requires an explicit calculation of the travel time sensitivity kernel,whereas the proposed method uses the wave-equation Born modeling framework combined with reverse time migration to reduce the computer memory and storage requirements.As such,it can be used for solving large-scale problems.Tests on numerical examples and field data demonstrated that the proposed method is able to invert high-resolution near-surface velocity models.