基于时变子波的品质因子估计

地下介质的地层滤波效应和吸收衰减效应会使地震信号频带变窄、主频降低、相位畸变。在对非稳态地震记录进行时变子波估计的基础上，从频率域和时间域两方面研究子波的时变性与品质因子的关系。首先推导非稳态记录正演模拟的时间域实现方法，通过子波褶积矩阵表达传播子波的时变性；其次，基于最优化思想，在给定范围内扫描Q值，使深部和浅部两个传播子波达到最佳匹配，提出了频率域子波匹配Q值估计方法和时间域子波匹配Q值估计方法。相比于常用的谱比法，频率域子波匹配法不需要做谱比，并通过最小二乘优化算法替代线性回归，具有更高的抗噪性。时间域子波匹配法通过子波褶积矩阵引入衰减响应，从根本上避免谱估计的误差，具有更高的精度。理论模型和实际数据计算结果表明，两种方法都能快速有效地估计品质因子，与时变子波谱比法相比，对低信噪比数据具有更强的鲁棒性。同时，模型测试表明品质因子估计结果的纵向分辨率有限，两个子波间的时间间隔制约着估计精度。

Quality factor Q estimation based on time-varying wavelet

The stratigraphic filtering and viscoelasticity attenuation can cause narrower seismic frequency band,lower dominant frequency,and phase distortion.The Q estimation and inverse Q filtering are important tools for attenuation compensation and resolution improvement.We propose a Q estimation methods for surface seismic datasets.The propagating wavelet is subject to stratigraphic filtering,viscoelastic attenuation,and concomitant dispersion,which causing time-varying characteristics.Firstly,the nonstationary seismic forward modeling is achieved in the time domain by wavelet convolve matrix,which shows the time-varying characteristics of propagating wavelet.Secondly,based on the estimation of time-varying propagating wavelets,we study the Q estimation from wavelets time-varying characteristics both in the time domain and frequency domain.The optimization method is used to scanning Qin order to minimize the difference between two wavelets at two different depths.The time-varying wavelet spectral match (TWSM) needs no spectral ratio or linear regression,therefore has better noise resistance than the logarithmic spectral ratio (LSR) method.The time-varying wavelet time-domain match (TWTM) has higher accuracy because of no spectral estimation.Tests on both synthetic and real data demonstrate that the two methods can achieve a fast Q estimation,and have better robust Q estimation on low SNR data than the time-varying wavelet logarithmic spectral ratio (TWLSR).The synthetic data examples also show relation between the Q estimation accuracy and the time interval of two wavelets,which means the Q estimation has resolution limitation.